ChatGPT [Investment opportunities in the Artificial Intelligence field]

Discuss all general (i.e. non-personal) investing questions and issues, investing news, and theory.
freakyfriday
Posts: 184
Joined: Tue Aug 23, 2022 10:00 am
Location: London

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by freakyfriday »

Graphics card manufacturers are doing exceptionally well.

Gamers can't get enough of them, crypto miners can't get enough, AI data centers can't get enough.

And each of them will buy someone else's surplus.
Whakamole
Posts: 1763
Joined: Wed Jan 13, 2016 8:59 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Whakamole »

freakyfriday wrote: Wed May 24, 2023 8:49 am Graphics card manufacturers are doing exceptionally well.

Gamers can't get enough of them, crypto miners can't get enough, AI data centers can't get enough.

And each of them will buy someone else's surplus.
This reminds me of Corning back in the dotcom days. I remember how no matter who won the Internet race, they'd need fiber, and how it was better to sell picks and shovels for gold miners.

GLW did do better than many dotcoms, since it's still in business, but it's under a third of the all-time high of $113, over twenty years later.
RJC
Posts: 1485
Joined: Fri Dec 14, 2018 12:40 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by RJC »

ThankYouJack wrote: Tue May 23, 2023 3:43 pm
PoorHomieQuan wrote: Tue May 23, 2023 2:46 pm
Valuethinker wrote: Tue May 23, 2023 3:30 am Google sees AI as an existential threat.
That's just a media narrative. Google has spent more money on AI research than any other company in history, and it didn't start doing that last year. Do yourself a favor and watch AlphaGo the Movie, if you haven't already. Google researchers invented LLMs, are leading in self-driving tech, and said AI more times than you can imagine at the last earnings call. What more could you want?!
As an investor, I would want to know:
1) how much revenue they anticipate losing in search (their bread and butter) because of AI
2) their revenue plan for AI
3) why Bard is (IMO) behind chatGPT if they’ve spent the most time and resources on AI

It’ll be interesting to see what happens. Just because google is pouring money into it doesn’t mean they’ll be the industry leader. they put a ton of resources into Facebook’s competitor, Google plus, and that failed miserably.
I use Bard more than ChatGTP-3 because the data is more current and I'm not interested in buying a subscription.
freakyfriday
Posts: 184
Joined: Tue Aug 23, 2022 10:00 am
Location: London

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by freakyfriday »

RJC wrote: Wed May 24, 2023 9:31 am
ThankYouJack wrote: Tue May 23, 2023 3:43 pm
PoorHomieQuan wrote: Tue May 23, 2023 2:46 pm
Valuethinker wrote: Tue May 23, 2023 3:30 am Google sees AI as an existential threat.
That's just a media narrative. Google has spent more money on AI research than any other company in history, and it didn't start doing that last year. Do yourself a favor and watch AlphaGo the Movie, if you haven't already. Google researchers invented LLMs, are leading in self-driving tech, and said AI more times than you can imagine at the last earnings call. What more could you want?!
As an investor, I would want to know:
1) how much revenue they anticipate losing in search (their bread and butter) because of AI
2) their revenue plan for AI
3) why Bard is (IMO) behind chatGPT if they’ve spent the most time and resources on AI

It’ll be interesting to see what happens. Just because google is pouring money into it doesn’t mean they’ll be the industry leader. they put a ton of resources into Facebook’s competitor, Google plus, and that failed miserably.
I use Bard more than ChatGTP-3 because the data is more current and I'm not interested in buying a subscription.
Bing Chat is effectively Microsoft's Bard for GPT-4.

The image generation on Bing is mindblowing, I believe it uses OpenAI's DALL-E but is much better than the API provided by OpenAI.
bg5
Posts: 562
Joined: Mon Jan 15, 2018 10:07 am

A.I is the next Tesla

Post by bg5 »

[Thread merged into here --admin LadyGeek]

Man....I am holding strong on the sidelines and sticking to my strategy but I have a gut feeling alot of money is gonna be made off this A.I. bubble.

Companies like Nvidia have almost 3x since October......hard to not be envious of the crazy returns of some of these investors but I keep trying to remind myself and slow and steady VTSAX wins the race or at least does pretty stinking well.

I just want Nvida type of returns :)
User avatar
watchnerd
Posts: 13609
Joined: Sat Mar 03, 2007 10:18 am
Location: Gig Harbor, WA, USA

Re: A.I is the next Tesla

Post by watchnerd »

The compute consumption is pretty subsidized right now.

Those cloud based GPUs suck up a lot of power.

When I was at megacorp, I could see how many cores OpenAI was using and it was a lot.
Global stocks, IG/HY bonds, gold & digital assets at market weights 75% / 19% / 6% || LMP: TIPS ladder
runner3081
Posts: 5978
Joined: Mon Aug 22, 2016 3:22 pm

Re: A.I is the next Tesla

Post by runner3081 »

I am confused... you already own Tesla and NVIDIA!?

VTSAX TOP 10 HOLDINGS
COMPANY SYMBOL TOTAL NET ASSETS
Apple Inc. AAPL 6.25%
Microsoft Corp. MSFT 5.61%
Amazon.com Inc. AMZN 2.25%
NVIDIA Corp. NVDA 1.59%
Alphabet Inc. Cl A GOOGL 1.57%
Berkshire Hathaway Inc. Cl B BRK.B 1.41%
Alphabet Inc. Cl C GOOG 1.34%
Meta Platforms Inc. META 1.31%
Exxon Mobil Corp. XOM 1.18%
UnitedHealth Group Inc. UNH 1.13%

Also in VTSAX.
Ticker Holdings Shares Market value
TSLA Tesla Inc. 83,576,682 $13,732,484,619
User avatar
LadyGeek
Site Admin
Posts: 95466
Joined: Sat Dec 20, 2008 4:34 pm
Location: Philadelphia
Contact:

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by LadyGeek »

I merged bg5's thread into the ongoing discussion.

(Thanks to the member who reported the post and provided a link to this thread.)
Wiki To some, the glass is half full. To others, the glass is half empty. To an engineer, it's twice the size it needs to be.
strummer6969
Posts: 978
Joined: Tue Dec 15, 2020 10:59 am

Re: A.I is the next Tesla

Post by strummer6969 »

watchnerd wrote: Thu May 25, 2023 8:40 am The compute consumption is pretty subsidized right now.

Those cloud based GPUs suck up a lot of power.

When I was at megacorp, I could see how many cores OpenAI was using and it was a lot.
I'm not a tech guy so probably a dumb question. But why not use TPUs? They're built specifically for machine learning tasks. It seems like TPU makers and fabs would be a better AI play than GPU makers.
investorpeter
Posts: 607
Joined: Sun Jul 31, 2016 5:46 pm

Re: A.I is the next Tesla

Post by investorpeter »

strummer6969 wrote: Thu May 25, 2023 9:12 am
watchnerd wrote: Thu May 25, 2023 8:40 am The compute consumption is pretty subsidized right now.

Those cloud based GPUs suck up a lot of power.

When I was at megacorp, I could see how many cores OpenAI was using and it was a lot.
I'm not a tech guy so probably a dumb question. But why not use TPUs? They're built specifically for machine learning tasks. It seems like TPU makers and fabs would be a better AI play than GPU makers.
Here’s what I said about Google’s TPU in 2017 (I think much of it still holds true):
investorpeter wrote: Sat Jun 17, 2017 10:43 pm I prefer Nvidia over Google as a stock right now. Google's tensor flow chip may be faster at certain applications, but Nvidia's GPUs are more adaptable to differing applications and I would argue that no one has come up with a transformative application for AI yet; autonomous driving being the leading candidate but still years away and I think that there will be even more significant applications to come that we have not even conceived yet. Nvidia has the lead in this field mong researchers since I can just go out and buy an Nvidia chip right now, whereas I cannot buy a Google tensor flow chip. Google has more money than Nvidia, but Nvidia still has the lead which is quite amazing when you think that not long ago, it's major market was video games. I like CEOs who can aggressively and successfully pivot companies into new industries (ala Reed Hastings at Netflix; or of course Bezos at Amazon, perhaps the master of the pivot). Google has not shown that it can pivot successfully away from search. Obviously, Nvidia has more room to grow than Google, even after its already meteoric rise. As a speculative stock investment (and all stock investments are speculative), I prefer Nvidia over Google at the moment. Though I still like Google very much. But I just see too many similarities between Google today and Microsoft in the 2000s. Google is also facing headwinds in Europe with all these regulatory agencies levying fines and such. Microsoft went through the same process in Europe and the U.S. and was significantly impaired by it. Google's corporate structure may be flawed, with too many things happening at once, without clear focus or leadership. Having said all of that, I still hedge my bets. I own Google, Amazon, Nvidia, and Microsoft.
strummer6969
Posts: 978
Joined: Tue Dec 15, 2020 10:59 am

Re: A.I is the next Tesla

Post by strummer6969 »

investorpeter wrote: Thu May 25, 2023 9:30 am
strummer6969 wrote: Thu May 25, 2023 9:12 am
watchnerd wrote: Thu May 25, 2023 8:40 am The compute consumption is pretty subsidized right now.

Those cloud based GPUs suck up a lot of power.

When I was at megacorp, I could see how many cores OpenAI was using and it was a lot.
I'm not a tech guy so probably a dumb question. But why not use TPUs? They're built specifically for machine learning tasks. It seems like TPU makers and fabs would be a better AI play than GPU makers.
Here’s what I said about Google’s TPU in 2017 (I think much of it still holds true):
investorpeter wrote: Sat Jun 17, 2017 10:43 pm I prefer Nvidia over Google as a stock right now. Google's tensor flow chip may be faster at certain applications, but Nvidia's GPUs are more adaptable to differing applications and I would argue that no one has come up with a transformative application for AI yet; autonomous driving being the leading candidate but still years away and I think that there will be even more significant applications to come that we have not even conceived yet. Nvidia has the lead in this field mong researchers since I can just go out and buy an Nvidia chip right now, whereas I cannot buy a Google tensor flow chip. Google has more money than Nvidia, but Nvidia still has the lead which is quite amazing when you think that not long ago, it's major market was video games. I like CEOs who can aggressively and successfully pivot companies into new industries (ala Reed Hastings at Netflix; or of course Bezos at Amazon, perhaps the master of the pivot). Google has not shown that it can pivot successfully away from search. Obviously, Nvidia has more room to grow than Google, even after its already meteoric rise. As a speculative stock investment (and all stock investments are speculative), I prefer Nvidia over Google at the moment. Though I still like Google very much. But I just see too many similarities between Google today and Microsoft in the 2000s. Google is also facing headwinds in Europe with all these regulatory agencies levying fines and such. Microsoft went through the same process in Europe and the U.S. and was significantly impaired by it. Google's corporate structure may be flawed, with too many things happening at once, without clear focus or leadership. Having said all of that, I still hedge my bets. I own Google, Amazon, Nvidia, and Microsoft.
Quite prescient; you called it. I remember how Nvidia totally disrupted CPU makers like Intel. For now at least, it doesn't seem there's much competition on the horizon. Nvidia is a great company but trading at many, many multiples of earnings.
LotsaGray
Posts: 1193
Joined: Sat Mar 25, 2023 2:08 pm

Re: ChatGPT

Post by LotsaGray »

Opinika wrote: Sun May 21, 2023 10:04 am A general comment about ChatGPT.

I gave it an assignment to answer some questions about meteorology, aerodynamics and physics. The same questions I would assign for sophomore college level final exams.

ChatGPT consistently aces my classes. Scary really. And so fast too.
Interesting. I have given it various questions. It does fine on the surface but when your start pushing it below the subject surface it fails badly. It will always present the answers very confidently even when answer in horribly wrong and often internally inconsistent. Most of these questions have been based on my engineering training from 40 yr ago. But some have been in general, nontechnical areas.

One of the most humorous was who the US senators from Louisiana were in 1931. Hint there was only one as Huey Long won in 1930 for the 1931-1937 term but did not swear in until mid 1932 ChatGPT had the basic facts right but could not put them together to realize one seat was vacant. First it was Long, th3n his predecessor, then was Longs wife (she filled his seat after he was shot in 1935), then was back to Long. It was obvious that the circle wasn’t going to stop unless I gave the answer. The wrong answers were all given with great confidence; almost a condescending tone.

It is easier to teach high #choolers than to lead ChatGPT to the right answer
Valuethinker
Posts: 48944
Joined: Fri May 11, 2007 11:07 am

Re: A.I is the next Tesla

Post by Valuethinker »

watchnerd wrote: Thu May 25, 2023 8:40 am The compute consumption is pretty subsidized right now.

Those cloud based GPUs suck up a lot of power.

When I was at megacorp, I could see how many cores OpenAI was using and it was a lot.
Aha!

Buy Electric Utilities!

:shock: :sharebeer
investorpeter
Posts: 607
Joined: Sun Jul 31, 2016 5:46 pm

Re: ChatGPT

Post by investorpeter »

LotsaGray wrote: Fri May 26, 2023 11:01 am
Opinika wrote: Sun May 21, 2023 10:04 am A general comment about ChatGPT.

I gave it an assignment to answer some questions about meteorology, aerodynamics and physics. The same questions I would assign for sophomore college level final exams.

ChatGPT consistently aces my classes. Scary really. And so fast too.
Interesting. I have given it various questions. It does fine on the surface but when your start pushing it below the subject surface it fails badly. It will always present the answers very confidently even when answer in horribly wrong and often internally inconsistent. Most of these questions have been based on my engineering training from 40 yr ago. But some have been in general, nontechnical areas.

One of the most humorous was who the US senators from Louisiana were in 1931. Hint there was only one as Huey Long won in 1930 for the 1931-1937 term but did not swear in until mid 1932 ChatGPT had the basic facts right but could not put them together to realize one seat was vacant. First it was Long, th3n his predecessor, then was Longs wife (she filled his seat after he was shot in 1935), then was back to Long. It was obvious that the circle wasn’t going to stop unless I gave the answer. The wrong answers were all given with great confidence; almost a condescending tone.

It is easier to teach high #choolers than to lead ChatGPT to the right answer
Agree. When asked very technical yes/no-type questions, ChatGPT sometimes gives the wrong answer. But it does so very confidently and even gives an explanation for the answer that would not necessarily be perceived as incorrect by someone without expert knowledge in that domain. It does better with open-ended opinion-type questions that can be answered with a long essay.
User avatar
CommitmentDevice
Posts: 272
Joined: Tue Apr 02, 2019 3:25 am

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by CommitmentDevice »

My hot take is that AI will be most valuable for startups.
Reason = AI makes it a lot cheaper to solve problems and create new products. It reduces incumbency advantage.

Take AudioPen.ai, for example. I just discovered this a few days ago. A developer could build it in a weekend. Yet, it solves a problem I have better than legacy systems so *snap* I'm now a user.

The counterfactual is that dominant players are also able to enhance their products using AI. Grammarly, for example, has a great AI writing assistant, for example. I'm sure we'll soon see something similar in Microsoft Word and Google Docs. So there won't be a strong incentive to switch to new platforms.
ThankYouJack
Posts: 5704
Joined: Wed Oct 08, 2014 7:27 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by ThankYouJack »

CommitmentDevice wrote: Fri May 26, 2023 12:35 pm My hot take is that AI will be most valuable for startups.
Reason = AI makes it a lot cheaper to solve problems and create new products. It reduces incumbency advantage.

Take AudioPen.ai, for example. I just discovered this a few days ago. A developer could build it in a weekend. Yet, it solves a problem I have better than legacy systems so *snap* I'm now a user.

The counterfactual is that dominant players are also able to enhance their products using AI. Grammarly, for example, has a great AI writing assistant, for example. I'm sure we'll soon see something similar in Microsoft Word and Google Docs. So there won't be a strong incentive to switch to new platforms.
Microsoft already announced 365 Copilot for https://blogs.microsoft.com/blog/2023/0 ... -for-work/

AI is incredible for development but I'm pretty sure AudioPen.ai would take longer than a weekend to build as there's a lot of behind the scenes stuff to consider.
jarjarM
Posts: 2502
Joined: Mon Jul 16, 2018 1:21 pm

Re: A.I is the next Tesla

Post by jarjarM »

watchnerd wrote: Thu May 25, 2023 8:40 am The compute consumption is pretty subsidized right now.

Those cloud based GPUs suck up a lot of power.

When I was at megacorp, I could see how many cores OpenAI was using and it was a lot.
As someone who's currently watching those AI racks getting deliver to the datacenter, the # of GPU cores being utilized for model training and inference is pretty outrageous. And the modeler are constantly demanding more cores and higher power (model training time is their metric). There's a huge reason why Nivida is seeing that much of an upside. Everyone is buying the H100s and in the 100ks per order too. Crazy how AWS/Azure/Google cloud all reduced their frontend web server and storage server but several X their request for GPU racks.
jarjarM
Posts: 2502
Joined: Mon Jul 16, 2018 1:21 pm

Re: A.I is the next Tesla

Post by jarjarM »

strummer6969 wrote: Thu May 25, 2023 9:12 am
watchnerd wrote: Thu May 25, 2023 8:40 am The compute consumption is pretty subsidized right now.

Those cloud based GPUs suck up a lot of power.

When I was at megacorp, I could see how many cores OpenAI was using and it was a lot.
I'm not a tech guy so probably a dumb question. But why not use TPUs? They're built specifically for machine learning tasks. It seems like TPU makers and fabs would be a better AI play than GPU makers.
Actually I asked the same question to the modelers, the answer is actually very simple. All of them were trained on GPU stack, so going to TPU and familiarized themself with TPU is a waster of time, especially given the fact that CUDA (Nvidia) has been around for 15+ years and a full library already deliver performance. Google's TPU library will require substantial tuning of the compiler and/or building a model on a new set of library. There's substantial pressure from leadership in all the big techs to deliver the model ASAP so no one want to waste their time right now (their metric is new model training time). That's why Nvidia is actually not fearful of a new competitor within the next 3-5 years. Unless the big 4 here in US band together like they did with OCP and really jointly develop something new with full set of widely adopted libraries.
hppycamper
Posts: 49
Joined: Sat May 22, 2021 8:41 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by hppycamper »

I wanted to dip my toes to invest in AI. Here's my story that I posted in another thread. BTW, it only impacts how small or large our children's inheritance will be. So I could take some risk.
hppycamper wrote: Thu May 25, 2023 8:26 pm At a seminar on creating an innovative culture back in early 2016, I heard Jensen speak. He shared then that an innovation idea (by a mathematician at NVIDIA!!) connected GPU to AI. He also shared his vision of optimizing GPU for AI, developing an SDK to make developing AI solutions using NVIDIA GPUs easy, and giving the SDK to developers for free. Then found this video https://www.youtube.com/watch?v=qhUvQiKec2U showcasing an example using the deep learning network powered by their chips. I was sold on his vision and got interested in buying shares of NVIDIA. It was around $10 per share at the time. Being a conservative investor, I didn't act on it, knowing the high risk of holding individual stocks. But I kept an eye on NVIDIA.

Fast forward to 2018, came across an article on AI startups. It said at the time, there were over 3000 AI startups and over 90% of them were using NVIDIA chips. At that point, I decided to invest my play money in AI via NVIDIA and bought at around $40 per share.

NVIDIA is the only stock I have bought. Today, I have almost 10X of my investment, all sitting in my Roth IRA. No regrets, only the lucky feeling :mrgreen:
User avatar
tinkertoyz
Posts: 36
Joined: Mon Dec 19, 2022 5:10 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by tinkertoyz »

Palantir

$PLTR
-tinkertoyz
Logan Roy
Posts: 1794
Joined: Sun May 29, 2022 10:15 am

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Logan Roy »

nisiprius wrote: Sun May 21, 2023 2:07 pm
carne_asada wrote: Sun May 21, 2023 12:33 pm
nisiprius wrote: Sun May 21, 2023 12:16 pm It seems like only yesterday (2011) that IBM's Watson beat Ken Jennings in Jeopardy! and everybody was saying all the same things about Watson.

So, how has IBM stock done since then?
IBM picked the wrong horse in the AI race and they quickly hit the limits of the the approach used for Jeopardy!
That's kind of the point. It's easy to say "AI" (or cannabis or aviation or personal computers or the Internet) has a bright future, but picking a broad category that everybody can see the potential of doesn't mean you don't have to pick the right horse. Picking the right horse is the hard part.
Absolutely. I think a large part of the hype behind ChatGPT is that it provides exponentially more value than it consumes. So the user is getting a lawyer, marketing consultant, virtual assistant, junior programmer, etc. and OpenAI are getting a few cents. So the only thing I'm sure of is human productivity is going to keep increasing for a while yet.

On the old subject of ChatGPT making things up, etc. there's probably the perfect explanation on the Jordan B. Peterson podcast, 360, from Dr Peter Attia.. Essentially ChatGPT isn't a knowledge engine – it's not researching anything; it doesn't have any resources; etc. It's a reasoning engine.. So all it's really doing is guessing what letter is most likely to come next in a sequence, that may be an answer to a question.

You could give it a knowledge base, and this would be a database, like wikipedia or the BMJ. But in effect it has no access to knowledge – just the echos of lots of information it was trained on in order to generate responses to things that demonstrate some kind of understanding.
User avatar
nisiprius
Advisory Board
Posts: 52105
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by nisiprius »

The core question is whether it is a good strategy to concentrate investments in narrow business categories if the category looks like a great opportunity--not whether AI is just such an opportunity.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
Logan Roy
Posts: 1794
Joined: Sun May 29, 2022 10:15 am

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Logan Roy »

nisiprius wrote: Sat May 27, 2023 12:11 pm The core question is whether it is a good strategy to concentrate investments in narrow business categories if the category looks like a great opportunity--not whether AI is just such an opportunity.
I'm hoping so, as I've recently moved nearly 20% into private infrastructure. Principally 3I Infrastructure, and BBGI Global Infrastructure.

These are moving to rare discounts, as yields are now as high, or higher, on cash. The market is a short-term optimiser – and it makes sense these investments should move to discounts. But if inflation and rates normalise, 20+ year 4-6% cashflows become attractive again, and if they don't, many of these contracts are inflation-linked and backed by government, with demand for infrastructure likely only increasing.

I don't care about beating the market – because I've no idea where the market's going to be – rates could hit 10%.. But I think concentrated bets can be a good strategy if the nature of the investment suits the investor. And with so much uncertainty, I quite like income at the moment.. One can also make the case that by only being invested in public stocks and bonds, the typical investor is probably quite underweight things like transport infrastructure, energy infrastructure, digitalisation, decarbonisation, etc. and making more concentrated bets on things like consumer electronics.
NabSh
Posts: 376
Joined: Fri Dec 25, 2020 11:09 am
Location: USA

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by NabSh »

If you know of a crystal ball that could predict which exact company would make a fortune then chances are so does everyone else too.

I say that predicting future of a company based on the past is extremely risky. I personally have lost money 2x chasing momentum stocks or past performance.

Having said that you may want keep less than 5 as play money or hobby. You still need to know when to get in and more importantly when to get out.
User avatar
burritoLover
Posts: 4097
Joined: Sun Jul 05, 2020 12:13 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by burritoLover »

This will be like the dot com bubble. The internet changed everything forever - it just didn't do it fast enough to realize actual earnings to justify sky-high valuations.
Logan Roy
Posts: 1794
Joined: Sun May 29, 2022 10:15 am

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Logan Roy »

NabSh wrote: Sun May 28, 2023 9:16 am If you know of a crystal ball that could predict which exact company would make a fortune then chances are so does everyone else too.
That's been the position of EMH for a long time. I'm not convinced it's the case.

I was watching a Peter Lynch interview from the 90s in which he was advocating retail investors stick to their area of expertise, and it's quite reasonable that someone who works in (e.g.) petrochemicals may know more about something than Wall Street.. That was 30 years into EMH – I'm not sure that's fundamentally changed over the following 30.

Not long ago, the combined valuations of the top tech growth stocks in the US would've necessitated the stock market nearly doubling in size just to fit them in at their projected levels of growth. I don't think it took geniuses to realise excess liquidity had driven things a bit too far. It was certainly easy for me not to buy 30 year bonds on negative real yields. Why would anyone want to secure negative real yields for 30 years while inflation was low??
User avatar
arcticpineapplecorp.
Posts: 15014
Joined: Tue Mar 06, 2012 8:22 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by arcticpineapplecorp. »

freakyfriday wrote: Mon May 22, 2023 3:27 am In my view the risk with a gold rush is that you invest in a mine and not in shovels.

OpenAI, Google, Meta etc are mine companies, AMD, Nvidia, Microsoft, Google, Amazon are shovel companies.

No matter who has the Next Great AI idea they will need to train the model on graphics cards in sever farms, and run it on (probably) graphics cards in server farms.
I don't understand. Is google a mine company or a shovel company? Or are you saying it's both?
It's hard to accept the truth when the lies were exactly what you wanted to hear. Investing is simple, but not easy. Buy, hold & rebalance low cost index funds & manage taxable events. Asking Portfolio Questions | Wiki
User avatar
dual
Posts: 1371
Joined: Mon Feb 26, 2007 6:02 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by dual »

Bbddl wrote: Sun May 21, 2023 9:27 am I was talking to somebody yesterday about the future of ChatGPT, and artificial intelligence. Anyone know of any investment opportunities in this field? Seems to have a very promising future…
There are still definitely problems with the technology. See this recent case about a lawyer using chatGPT to prepare a legal filing. The software made up citations. The fake citations were caught by the opposing counsel who brought them to the attention of the judge. He was not amused.

https://www.nytimes.com/2023/05/27/nyre ... atgpt.html

Here’s What Happens When Your Lawyer Uses ChatGPT
A lawyer representing a man who sued an airline relied on artificial intelligence to help prepare a court filing. It did not go well.
freakyfriday
Posts: 184
Joined: Tue Aug 23, 2022 10:00 am
Location: London

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by freakyfriday »

arcticpineapplecorp. wrote: Sun May 28, 2023 9:43 am
freakyfriday wrote: Mon May 22, 2023 3:27 am In my view the risk with a gold rush is that you invest in a mine and not in shovels.

OpenAI, Google, Meta etc are mine companies, AMD, Nvidia, Microsoft, Google, Amazon are shovel companies.

No matter who has the Next Great AI idea they will need to train the model on graphics cards in sever farms, and run it on (probably) graphics cards in server farms.
I don't understand. Is google a mine company or a shovel company? Or are you saying it's both?
Google are all three, they have data centres (of course) but they also make AI tech like the Edge TPU & they also research AI extensively. They are also extensive consumers of AI.

In fact Google kicked off the modern large language model era.

Microsoft of course has a large stake in OpenAI so it bridges that gap.

Apple now designs it's own processors which have extensive AI capabilities.
User avatar
nisiprius
Advisory Board
Posts: 52105
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by nisiprius »

dual wrote: Sun May 28, 2023 10:20 am There are still definitely problems with the technology. See this recent case about a lawyer using chatGPT to prepare a legal filing. The software made up citations. The fake citations were caught by the opposing counsel who brought them to the attention of the judge. He was not amused.

https://www.nytimes.com/2023/05/27/nyre ... atgpt.html...
I think this is important, not just one isolated glitch. In the current state of things (I can't deal with unverifiable assertions that all the problems will be solved six months from now) you dare not use ChatGPT or the others for anything important, unless you are going to have a human fact-check the results. And that costs something. It costs something either way: in lawsuits against you if you don't, and in the cost of checking if you do.

How much of the cost savings from using these models will get eaten up by the cost of fact-checking them?

Information that is 90% correct doesn't have 90% of the value of correct information.

Most disturbing is that these models do not seem to have the capability of doing their own fact-checking, and they don't seem to know the sources of their information. I suspect that sources actually get lost in the training process, and all that is left of the original information is the traces left in the new neuronal connections... generalizations drawn from the original. Ask ChatGPT for a source, the truth is that all it knows is "I think I once read somewhere." So what does ChatGPT do when asked for a source? It makes a plausible guess as to a place where you'd expect to find the information, and calls it the source--without checking it.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
cacophony
Posts: 1354
Joined: Tue Oct 16, 2007 9:12 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by cacophony »

nisiprius wrote: Sun May 28, 2023 11:23 am
dual wrote: Sun May 28, 2023 10:20 am There are still definitely problems with the technology. See this recent case about a lawyer using chatGPT to prepare a legal filing. The software made up citations. The fake citations were caught by the opposing counsel who brought them to the attention of the judge. He was not amused.

https://www.nytimes.com/2023/05/27/nyre ... atgpt.html...
I think this is important, not just one isolated glitch. In the current state of things (I can't deal with unverifiable assertions that all the problems will be solved six months from now) you dare not use ChatGPT or the others for anything important, unless you are going to have a human fact-check the results. And that costs something. It costs something either way: in lawsuits against you if you don't, and in the cost of checking if you do.

How much of the cost savings from using these models will get eaten up by the cost of fact-checking them?

....
As many are already discovering, there are a lot of situations where the work it's able to generate produces far more value than the costs associated with fact checking. In the case of writing or coding for example, you can tell within seconds whether the response will be of use.
Yarlonkol12
Posts: 968
Joined: Thu Apr 11, 2019 4:28 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Yarlonkol12 »

Bbddl wrote: Sun May 21, 2023 9:27 am I was talking to somebody yesterday about the future of ChatGPT, and artificial intelligence. Anyone know of any investment opportunities in this field? Seems to have a very promising future…
I think the opportunity is probably avoiding any "AI" focused investments that are currently being offered, if you are able to do that, you will save yourself a lot of money when the bubble pops and the talking heads switch over to the next fad
My posts are for entertainment purposes only.
User avatar
dual
Posts: 1371
Joined: Mon Feb 26, 2007 6:02 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by dual »

coding for example, you can tell within seconds whether the response will be of use.
I have written a lot of code, and that is not my experience at all. Code has to be tested thoroughly. A difficult task: every branch and function has to be exercised and limiting cases tested.

Here’s an interesting example of a software failure. The Japanese Hakuto-R moon lander crashed and the development team is doing a post Mortem. Here’s what they found.
ispace engineers discovered that there was a conflict between these two sources of data: the estimated altitude over the Moon and the distance reported by the laser rangefinder recorded a difference of about 5 km (3.1 miles).

What happened then? The moonlander ignored the rangefinder data, burning its propellant too quickly and spending two minutes in free fall.

ispace chief technology officer Ryo Ujiie noted that this fatal discrepancy came from a last minute decision to move the robot’s target landing site from a flat plain on the Moon called Lacus Somniorum to the Atlas crater.
https://interestingengineering.com/scie ... n-the-moon

The code was NOT Generated by AI. I’m just giving an example of possible problems with testing software.
cacophony
Posts: 1354
Joined: Tue Oct 16, 2007 9:12 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by cacophony »

dual wrote: Sun May 28, 2023 1:23 pm
coding for example, you can tell within seconds whether the response will be of use.
I have written a lot of code, and that is not my experience at all. Code has to be tested thoroughly. A difficult task: every branch and function has to be exercised and limiting cases tested.

Here’s an interesting example of a software failure. The Japanese Hakuto-R moon lander crashed and the development team is doing a post Mortem. Here’s what they found.
ispace engineers discovered that there was a conflict between these two sources of data: the estimated altitude over the Moon and the distance reported by the laser rangefinder recorded a difference of about 5 km (3.1 miles).

What happened then? The moonlander ignored the rangefinder data, burning its propellant too quickly and spending two minutes in free fall.

ispace chief technology officer Ryo Ujiie noted that this fatal discrepancy came from a last minute decision to move the robot’s target landing site from a flat plain on the Moon called Lacus Somniorum to the Atlas crater.
https://interestingengineering.com/scie ... n-the-moon

The code was NOT Generated by AI. I’m just giving an example of possible problems with testing software.
I'm talking about more general help. Higher level uses for ideas/approaches or simply as a generation tool for structural elements, mock data, etc. There are also many cases where you can get a solution that is 90% of the way there and fixing the last 10% is a lot easier than doing it all from scratch. Obviously nobody should use it to generate the entirety of mission critical software.
Logan Roy
Posts: 1794
Joined: Sun May 29, 2022 10:15 am

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Logan Roy »

nisiprius wrote: Sun May 28, 2023 11:23 am
dual wrote: Sun May 28, 2023 10:20 am There are still definitely problems with the technology. See this recent case about a lawyer using chatGPT to prepare a legal filing. The software made up citations. The fake citations were caught by the opposing counsel who brought them to the attention of the judge. He was not amused.

https://www.nytimes.com/2023/05/27/nyre ... atgpt.html...
I think this is important, not just one isolated glitch. In the current state of things (I can't deal with unverifiable assertions that all the problems will be solved six months from now) you dare not use ChatGPT or the others for anything important, unless you are going to have a human fact-check the results. And that costs something. It costs something either way: in lawsuits against you if you don't, and in the cost of checking if you do.

How much of the cost savings from using these models will get eaten up by the cost of fact-checking them?

Information that is 90% correct doesn't have 90% of the value of correct information.

Most disturbing is that these models do not seem to have the capability of doing their own fact-checking, and they don't seem to know the sources of their information. I suspect that sources actually get lost in the training process, and all that is left of the original information is the traces left in the new neuronal connections... generalizations drawn from the original. Ask ChatGPT for a source, the truth is that all it knows is "I think I once read somewhere." So what does ChatGPT do when asked for a source? It makes a plausible guess as to a place where you'd expect to find the information, and calls it the source--without checking it.
Again though, ChatGPT is not a 'knowledge engine' .. A knowledge engine would be something that access to a large vector of recorded data – like wikipedia.

ChatGPT is a 'reasoning engine'. So it's got no access to any data whatsoever. All it's doing is predicting what the most likely next letter/word should be in a sequence. Whatever knowledge/experience it appears to have is just the result of having learnt to predict word sequences across vast amounts of data – which will itself be as unreliable as the humans that wrote it.

You can quite easily hook ChatGPT up to a database, encyclopaedia, etc. That's not what's impressive about what it's doing. Answering quiz questions just requires something that can look up indexes.
User avatar
dual
Posts: 1371
Joined: Mon Feb 26, 2007 6:02 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by dual »

You can quite easily hook ChatGPT up to a database, encyclopaedia, etc. That's not what's impressive about what it's doing. Answering quiz questions just requires something that can look up indexes.
People certainly use chatGPT to answer questions and they do (foolishly) trust the results. Researchers in AI have known about hallucinations since at least 2018. Why has it not been fixed?

Some researchers say it’s a fundamental problem that cannot be fixed.
But Yann LeCun, a pioneer in deep learning and the self-supervised learning used in large language models, believes there is a more fundamental flaw that leads to hallucinations.

“Large language models have no idea of the underlying reality that language describes,” he said, adding that most human knowledge is nonlinguistic. “Those systems generate text that sounds fine, grammatically, semantically, but they don’t really have some sort of objective other than just satisfying statistical consistency with the prompt.”

Humans operate on a lot of knowledge that is never written down, such as customs, beliefs, or practices within a community that are acquired through observation or experience. And a skilled craftsperson may have tacit knowledge of their craft that is never written down.

“Language is built on top of a massive amount of background knowledge that we all have in common, that we call common sense,” LeCun said. He believes that computers need to learn by observation to acquire this kind of nonlinguistic knowledge.

“There is a limit to how smart they can be and how accurate they can be because they have no experience of the real world, which is really the underlying reality of language,” said LeCun. “Most of what we learn has nothing to do with language.”

https://spectrum.ieee.org/ai-hallucination
User avatar
nisiprius
Advisory Board
Posts: 52105
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by nisiprius »

Logan Roy wrote: Sun May 28, 2023 3:52 pmAgain though, ChatGPT is not a 'knowledge engine' ...
Maybe not, but it is being presented as such and used as such right now. And people are investing in it with the idea that that's that's what it is.

Nobody put a gun to Microsoft's head and said "have it open the conversation with 'ask me anything.' Nobody put a gun to Forefront's head and said "be sure to give all the response profiles names and portraits."
A knowledge engine would be something that access to a large vector of recorded data – like wikipedia...
I thought I had read that Wikipedia was in fact part of the training corpus. Indeed:...12.5% comes from a public dataset of crawled content from the web and another 12.5% comes from Wikipedia."
You can quite easily hook ChatGPT up to a database, encyclopaedia, etc.
If it's quite easy, they should have done it before exposing it to the general public.
That's not what's impressive about what it's doing. Answering quiz questions just requires something that can look up indexes.
This thread is about concentrated investing in AI companies in hope of making money. It shouldn't be about "impressive," it should be about "valuable--in dollars."

If ChatGPT, Bing Chat, and Bard are not knowledge engines, but just brainstorming partners, let them say so. If they are just a way to generate crazy ideas to challenge ossified thinking, and if they do it better than a $16.99 Creative Whack Pack card deck or I Ching, let them say so.

When Steven A. Schwartz of Levidow, Levidow & Oberman let ChatGPT generate a brief and submitted it to a court of law, he wasn't using it for creative inspiration. He thought it was a knowledge engine producing accurate finished work. Why? Because that's how it is being presented, that's why.
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
User avatar
nisiprius
Advisory Board
Posts: 52105
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by nisiprius »

And if ChatGPT, Bard, and Bing Chat are "not AI..." let them say so.

Image

Image

Image
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
User avatar
Hacksawdave
Posts: 784
Joined: Tue Feb 14, 2023 4:44 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Hacksawdave »

I was in Systems and Data Processing (IT) for nearly 40 years and introduced to AI back in 1986-1988. Sorry, it’s not something new. I’ve seen many ‘new’ technologies the last 12 years that are basically rehashes of prior ones. VM, Big Data, and cloud were already done in the 1970s.

One poster said the best way is through VTSAX, and that I agree upon. If the technology really is a true innovation, it will make it to the big cap funds like VTSAX and VFINX. I own both and one just must have patience.
User avatar
SimpleGift
Posts: 4477
Joined: Tue Feb 08, 2011 2:45 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by SimpleGift »

dual wrote: Sun May 28, 2023 4:32 pm Some researchers say it’s a fundamental problem that cannot be fixed.
But Yann LeCun, a pioneer in deep learning and the self-supervised learning used in large language models, believes there is a more fundamental flaw that leads to hallucinations.

“Large language models have no idea of the underlying reality that language describes,” he said, adding that most human knowledge is nonlinguistic. “Those systems generate text that sounds fine, grammatically, semantically, but they don’t really have some sort of objective other than just satisfying statistical consistency with the prompt.”

Humans operate on a lot of knowledge that is never written down, such as customs, beliefs, or practices within a community that are acquired through observation or experience. And a skilled craftsperson may have tacit knowledge of their craft that is never written down.

“Language is built on top of a massive amount of background knowledge that we all have in common, that we call common sense,” LeCun said. He believes that computers need to learn by observation to acquire this kind of nonlinguistic knowledge.

“There is a limit to how smart they can be and how accurate they can be because they have no experience of the real world, which is really the underlying reality of language,” said LeCun. “Most of what we learn has nothing to do with language.”
https://spectrum.ieee.org/ai-hallucination
Thank you for posting this penetrating insight by Mr. LeCun regarding the limitations of these large language models.

Not sure many folks grasp that they are limited to linguistic knowledge only, which is not tempered or qualified by any wisdom or experience from the non-linguistic world. On the one hand, this is reassuring (i.e., in their current form, they won't be supplanting important human roles anytime soon) — but on the other hand, the misunderstanding or misapplication of their limited capabilities appears even more dangerous.
Logan Roy
Posts: 1794
Joined: Sun May 29, 2022 10:15 am

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Logan Roy »

nisiprius wrote: Sun May 28, 2023 4:51 pm
Logan Roy wrote: Sun May 28, 2023 3:52 pmAgain though, ChatGPT is not a 'knowledge engine' ...
Maybe not, but it is being presented as such and used as such right now. And people are investing in it with the idea that that's that's what it is.

Nobody put a gun to Microsoft's head and said "have it open the conversation with 'ask me anything.' Nobody put a gun to Forefront's head and said "be sure to give all the response profiles names and portraits."
A knowledge engine would be something that access to a large vector of recorded data – like wikipedia...
I thought I had read that Wikipedia was in fact part of the training corpus. Indeed:...12.5% comes from a public dataset of crawled content from the web and another 12.5% comes from Wikipedia."
You can quite easily hook ChatGPT up to a database, encyclopaedia, etc.
If it's quite easy, they should have done it before exposing it to the general public.
That's not what's impressive about what it's doing. Answering quiz questions just requires something that can look up indexes.
This thread is about concentrated investing in AI companies in hope of making money. It shouldn't be about "impressive," it should be about "valuable--in dollars."

If ChatGPT, Bing Chat, and Bard are not knowledge engines, but just brainstorming partners, let them say so. If they are just a way to generate crazy ideas to challenge ossified thinking, and if they do it better than a $16.99 Creative Whack Pack card deck or I Ching, let them say so.

When Steven A. Schwartz of Levidow, Levidow & Oberman let ChatGPT generate a brief and submitted it to a court of law, he wasn't using it for creative inspiration. He thought it was a knowledge engine producing accurate finished work. Why? Because that's how it is being presented, that's why.
It's trained on a huge volume of work, but it doesn't specifically store or index any of it, nor does it prioritise accuracy. The information exists as echos and fragments within the connections it uses to predict the next letter in a sequence.

The fact a lot of people don't realise what it is and what it's for says something about society – I think AI researchers have been quite clear on this. This podcast with Brian Roemmelle explain it very clearly:
https://www.jordanbpeterson.com/podcast ... le-ep-357/

The other big part of this is that results you can get from ChatGPT can be made 900% more accurate or intelligent through 'prompt engineering'. So the way you pose it a question can have a huge effect on how it thinks a problem out. And the current state-of-the-art method is called the Tree of Thought Principle.. In many ways, asking it face value questions is like talking to a drunk college professor – it recognises sentences, and words come out of its mouth. But through logically posing questions and insisting on thought processes, it can solve crossword puzzles and creative writing tasks at a much higher level – being forced to think things through, work step-by-step, etc. It's a much bigger breakthrough than 99% of people realise at the moment.
ThankYouJack
Posts: 5704
Joined: Wed Oct 08, 2014 7:27 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by ThankYouJack »

Logan Roy wrote: Sun May 28, 2023 6:21 pm It's a much bigger breakthrough than 99% of people realise at the moment.
I'm not sure about 99% but I think there is a big disconnect between the general public's understanding and why I would invest directly in OpenAI if I could. We have very smart Bogleheads thinking it's a fad / too inaccurate to be useful / over-hyped / maybe even pointless but if that's the case why would tech companies be investing literally tens of billions of dollars into the technology?
User avatar
dual
Posts: 1371
Joined: Mon Feb 26, 2007 6:02 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by dual »

ThankYouJack wrote: Mon May 29, 2023 11:37 am
Logan Roy wrote: Sun May 28, 2023 6:21 pm It's a much bigger breakthrough than 99% of people realise at the moment.
I'm not sure about 99% but I think there is a big disconnect between the general public's understanding and why I would invest directly in OpenAI if I could. We have very smart Bogleheads thinking it's a fad / too inaccurate to be useful / over-hyped / maybe even pointless but if that's the case why would tech companies be investing literally tens of billions of dollars into the technology ?
Billions have also been invested in virtual reality mainly by Facebook with no payoff. With research and development you pay your money and you take your chances. I think the excitement over large language models will fade once people become aware of its very real problems mentioned in this thread.
cacophony
Posts: 1354
Joined: Tue Oct 16, 2007 9:12 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by cacophony »

dual wrote: Mon May 29, 2023 1:03 pm
ThankYouJack wrote: Mon May 29, 2023 11:37 am
Logan Roy wrote: Sun May 28, 2023 6:21 pm It's a much bigger breakthrough than 99% of people realise at the moment.
I'm not sure about 99% but I think there is a big disconnect between the general public's understanding and why I would invest directly in OpenAI if I could. We have very smart Bogleheads thinking it's a fad / too inaccurate to be useful / over-hyped / maybe even pointless but if that's the case why would tech companies be investing literally tens of billions of dollars into the technology ?
Billions have also been invested in virtual reality mainly by Facebook with no payoff. With research and development you pay your money and you take your chances. I think the excitement over large language models will fade once people become aware of its very real problems mentioned in this thread.
There's a big difference between one individual at one company investing billions in a project idea that nobody else thought made any sense, and most industries making a shift and investment. The reality is that many industries are already seeing noticeable benefits from the current state of generative AI and we are certain to see the impacts of this over the next few years. The only real question at this point is whether the impacts will be noticeable in certain industries, massive across all industries, or somewhere in between.

The problems mentioned in this thread are widely known and easily discoverable. At this point I think of it akin to a craftsman blaming his tools. Prompt engineering is becoming a thing for a reason.
User avatar
nisiprius
Advisory Board
Posts: 52105
Joined: Thu Jul 26, 2007 9:33 am
Location: The terrestrial, globular, planetary hunk of matter, flattened at the poles, is my abode.--O. Henry

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by nisiprius »

dual wrote: Sun May 28, 2023 4:32 pm ... https://spectrum.ieee.org/ai-hallucination ...
Thank you. That article led me to this one:
Medical image AIs need a good "hallucination map" My underlining:
medical imaging devices do not record images directly. Instead, the raw data collected by the devices is analyzed by a computer, and machine-learning algorithms are used to reconstruct the images that doctors and radiologists use for diagnosing a health complication. Image reconstruction is done based on the known physics of the imaging device, along with a set of assumptions about how the final image should appear.

“However, if certain assumptions are wrong during image reconstruction, false structures may be introduced into the final image,” explains Mark Anastasio, a professor of bioengineering at the University of Illinois at Urbana–Champaign.

As one could imagine, the presence of false structures (also called “hallucinations”) in medical images could have serious consequences in diagnosing patients.
In other words, AI is already introducing hallucinations into medical images. (The article is, of course, about solving this by introducing yet more technology--an "algorithm," not clear if it's AI-based or not, that guesses what structures seen in images might be fake).
Annual income twenty pounds, annual expenditure nineteen nineteen and six, result happiness; Annual income twenty pounds, annual expenditure twenty pounds ought and six, result misery.
User avatar
dual
Posts: 1371
Joined: Mon Feb 26, 2007 6:02 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by dual »

Cacophony
The problems mentioned in this thread are widely known and easily discoverable. At this point I think of it akin to a craftsman blaming his tools. Prompt engineering is becoming a thing for a reason.
Easily discernible does not mean they are easily solved.

As shown in Nisiprius post above, the AI companies are selling these as solutions to problems that require accuracy. So the craftsman is falsely selling a product of his tools

Yes, there is always some magic fix like “prompt engineering” that is sold. But that does not mean it works. As far as I know no system with “prompt engineering” has been widely deployed. It is not used in the widely available chat bots.
cacophony
Posts: 1354
Joined: Tue Oct 16, 2007 9:12 pm

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by cacophony »

dual wrote: Mon May 29, 2023 3:34 pm Cacophony
The problems mentioned in this thread are widely known and easily discoverable. At this point I think of it akin to a craftsman blaming his tools. Prompt engineering is becoming a thing for a reason.
Easily discernible does not mean they are easily solved.

As shown in Nisiprius post above, the AI companies are selling these as solutions to problems that require accuracy. So the craftsman is falsely selling a product of his tools

Yes, there is always some magic fix like “prompt engineering” that is sold. But that does not mean it works. As far as I know no system with “prompt engineering” has been widely deployed. It is not used in the widely available chat bots.
My point is that even without solving any of the current problems we are already in a state that will achieve far more than the prediction that it will be a "fad / too inaccurate to be useful / over-hyped / maybe even pointless".

How these technologies are marketed and sold is an entirely different discussion topic.

Prompt engineering is something that you the user does:

"Prompt engineering is a relatively new discipline for developing and optimizing prompts to efficiently use language models (LMs) for a wide variety of applications and research topics. Prompt engineering skills help to better understand the capabilities and limitations of large language models (LLMs)."

There are plenty of resources trying to help teach this and many companies are now hiring people with these skills:
https://www.google.com/search?q=prompt+engineering+jobs

Put simply, how you ask the question can make a big difference in terms of the quality of the response.
Logan Roy
Posts: 1794
Joined: Sun May 29, 2022 10:15 am

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by Logan Roy »

ThankYouJack wrote: Mon May 29, 2023 11:37 am
Logan Roy wrote: Sun May 28, 2023 6:21 pm It's a much bigger breakthrough than 99% of people realise at the moment.
I'm not sure about 99% but I think there is a big disconnect between the general public's understanding and why I would invest directly in OpenAI if I could. We have very smart Bogleheads thinking it's a fad / too inaccurate to be useful / over-hyped / maybe even pointless but if that's the case why would tech companies be investing literally tens of billions of dollars into the technology?
I think specifically, GPT is one of those leaps in technology you can only compare to the invention of the microprocessor and Internet.. Investors chase lots of hyped ideas – I still don't know whether we'll find a real use for blockchain – but GPT is basically Artificial General Intelligence – or at least something very close to it.. And like the Internet and microprocessor, you can apply this technology to almost anything, and it transforms that thing.

Put it in a virtual assistant – like Siri or Alexa – and you've got an omnipresent consultant, doctor, therapist, advisor, virtual friend, lecturer, employee.. You've got something that could be a clone of you, and could go and do things on your behalf.. It's all doable now – and it'll take ordinary people years to wake up to how revolutionary this is.. It might be more comparable to the industrial revolution – the Internet and microprocessor may have just been stepping stones to this. We had HAL9000 in mind when both of these technologies were in foetal stages.

I'm not sure there's anything specifically investable. I think the biggest effect will be on human productivity growth. So it'll be stocks, but I suppose there are too many applications for me to imagine any particular sector leaping ahead. I do think it'll remove the equity risk premium – because AI is going to become the master capital allocator, and you won't really need competitive trading anymore, while risk in markets should be a lot lower.
jebmke
Posts: 25271
Joined: Thu Apr 05, 2007 2:44 pm
Location: Delmarva Peninsula

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by jebmke »

dual wrote: Mon May 29, 2023 1:03 pm
ThankYouJack wrote: Mon May 29, 2023 11:37 am
Logan Roy wrote: Sun May 28, 2023 6:21 pm It's a much bigger breakthrough than 99% of people realise at the moment.
I'm not sure about 99% but I think there is a big disconnect between the general public's understanding and why I would invest directly in OpenAI if I could. We have very smart Bogleheads thinking it's a fad / too inaccurate to be useful / over-hyped / maybe even pointless but if that's the case why would tech companies be investing literally tens of billions of dollars into the technology ?
Billions have also been invested in virtual reality mainly by Facebook with no payoff. With research and development you pay your money and you take your chances. I think the excitement over large language models will fade once people become aware of its very real problems mentioned in this thread.
Right; the Metaverse was going to break everything wide open. How much capital was destroyed in that bet?
Stay hydrated; don't sweat the small stuff
adave
Posts: 549
Joined: Thu Sep 13, 2007 7:17 pm
Location: Houston

Re: ChatGPT [Investment opportunities in the Artificial Intelligence field]

Post by adave »

ChatGPT is easily fooled by capital letters.

This tech is a big deal sure but has many issues that may never be resolved.

Problem is nobody really understands how it works.
Locked