New google AI called Bard

Questions on how we spend our money and our time - consumer goods and services, home and vehicle, leisure and recreational activities
User avatar
Topic Author
heartwood
Posts: 2700
Joined: Sat Nov 23, 2013 12:40 pm

New google AI called Bard

Post by heartwood »

This is a follow up to the Bing AI thread. I cross posted there. No joy for me with Bing!


I just got invited and accepted to the new google Bard, it's new conversational AI.

It seems to work for me out of the box.

Here's it's response to my send a birthday card to chairman mao from general macarthur.

Happy birthday card from Douglas MacArthur to chairman mao


Dear Chairman Mao,

Happy birthday! I hope you have a wonderful day filled with joy and happiness.

I am writing to you today as a former adversary and now a friend. I know that we have had our differences in the past, but I believe that we can now work together to build a better future for our countries.

I have always admired your leadership and your vision for China. You have brought your country from a state of poverty and chaos to a state of prosperity and stability. I believe that you can do even more in the years to come.

I hope that you will accept this card as a token of my goodwill. I wish you all the best for the future.

Sincerely,

Douglas MacArthur





You can find Bard at https://bard.google.com/

Perhaps a moderator can decide if we want one AI conversational thread for all engines, or specific threads for the major public ones.
mega317
Posts: 5705
Joined: Tue Apr 19, 2016 10:55 am

Re: New google AI called Bard

Post by mega317 »

Wow seems like revolutionary technology there.
Try asking it some of the questions that were answered incorrectly in other threads.
hbdad
Posts: 239
Joined: Fri Dec 13, 2019 2:29 pm

Re: New google AI called Bard

Post by hbdad »

heartwood wrote: Wed Mar 22, 2023 12:46 pm This is a follow up to the Bing AI thread. I cross posted there. No joy for me with Bing!


I just got invited and accepted to the new google Bard, it's new conversational AI.

It seems to work for me out of the box.

Here's it's response to my send a birthday card to chairman mao from general macarthur.

Happy birthday card from Douglas MacArthur to chairman mao


Dear Chairman Mao,

Happy birthday! I hope you have a wonderful day filled with joy and happiness.

I am writing to you today as a former adversary and now a friend. I know that we have had our differences in the past, but I believe that we can now work together to build a better future for our countries.

I have always admired your leadership and your vision for China. You have brought your country from a state of poverty and chaos to a state of prosperity and stability. I believe that you can do even more in the years to come.

I hope that you will accept this card as a token of my goodwill. I wish you all the best for the future.

Sincerely,

Douglas MacArthur





You can find Bard at https://bard.google.com/

Perhaps a moderator can decide if we want one AI conversational thread for all engines, or specific threads for the major public ones.
This Bard must be supported by the CCP. Somehow I doubt General MacArthur would ever write those words.
mkmal1
Posts: 27
Joined: Mon Dec 03, 2007 11:14 am

Re: New google AI called Bard

Post by mkmal1 »

If you listen to Bard you can save a lot less for retirement :D
How much could you take out of a 1 million dollar portolio set at 60 percent stocks 30 percent bonds, 10 percent cash anually and never deplete the inflation adjusted starting value?

You can withdraw $72,000 annually from a $1 million portfolio set at 60% stocks, 30% bonds, and 10% cash and never deplete the inflation adjusted starting value.

To calculate this, we need to first determine the expected rate of return on the portfolio. We can do this by using the following formula:

Expected Rate of Return = (% Stocks * Expected Return on Stocks) + (% Bonds * Expected Return on Bonds) + (% Cash * Expected Return on Cash)

Let's assume that the expected return on stocks is 10%, the expected return on bonds is 5%, and the expected return on cash is 0%. This gives us an expected rate of return of 7.5%.

Next, we need to determine the inflation rate. We can do this by looking at the historical inflation rate or by using a forecast from a reputable source. For this example, let's assume that the inflation rate is 3%.

Now, we can calculate the amount that can be withdrawn annually by using the following formula:

Code snippet
Withdrawal Amount = (Portfolio Value * Inflation Rate) / (1 + Expected Rate of Return)
Use code with caution. Learn more
Plugging in the numbers from our example, we get:

Code snippet
Withdrawal Amount = ($1,000,000 * 0.03) / (1 + 0.075)
= $72,000
Use code with caution. Learn more
This means that you can withdraw $72,000 annually from a $1 million portfolio set at 60% stocks, 30% bonds, and 10% cash and never deplete the inflation adjusted starting value.

It is important to note that this is just an estimate. The actual amount that you can withdraw will vary depending on the actual performance of your investments.
jebmke
Posts: 25475
Joined: Thu Apr 05, 2007 2:44 pm
Location: Delmarva Peninsula

Re: New google AI called Bard

Post by jebmke »

mkmal1 wrote: Sat May 06, 2023 5:52 pm If you listen to Bard you can save a lot less for retirement :D

Isn't one of the issues with these AI machines hallucinations? :mrgreen:
Don't trust me, look it up. https://www.irs.gov/forms-instructions-and-publications
cacophony
Posts: 1363
Joined: Tue Oct 16, 2007 9:12 pm

Re: New google AI called Bard

Post by cacophony »

Here's how GPT-4 handled it:
Dear Chairman Mao,

On this special day, I, General MacArthur, extend my warmest birthday wishes to you. Though our paths have diverged and our ideologies differ, I recognize the significance of this occasion and the impact you have had on your nation.

May this day bring you joy and reflection, and may the future hold opportunities for understanding and cooperation between our peoples.

Sincerely,
General Douglas MacArthur
User avatar
arcticpineapplecorp.
Posts: 15080
Joined: Tue Mar 06, 2012 8:22 pm

Re: New google AI called Bard

Post by arcticpineapplecorp. »

Mao's Great Leap Forward was responsible for at least 30 Million People dying of starvation "more than any other single famine in recorded human history. Most tragically, this disaster was largely preventable."

It was Deng Xiaoping's policies that paved the way for the transformation that we now see in China.

Another epic fail for Bard I believe, unless that is, General MacArthur was a natural born politician.
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
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

Q: What is the surface area of a pyramid 12 meters high, standing on a base that is a 10-meter square?
Bard wrote:The surface area of a pyramid is the sum of the areas of its faces. In this case, the pyramid has a square base with sides of length 10 meters and a height of 12 meters. The four lateral faces of the pyramid are triangles. Each triangle has a base of length 10 meters and a height of 12 meters. The area of a triangle is equal to half the product of its base and height, so the area of each lateral face is (1/2)(10)(12) = 60 square meters. The total area of the four lateral faces is 60 * 4 = 240 square meters. The area of the square base is 10 * 10 = 100 square meters. The total surface area of the pyramid is 240 + 100 = 340 square meters.

Therefore, the surface area of the pyramid is 340 square meters.
Serious fail. The area of each triangle is half the product of its base and altitude, but the triangular faces are on a slant. The height of each triangle is not the 12 meter height of the pyramid, it is the the hypotenuse of a 5-12-13 right triangle and is 13 meters. Each lateral face has an area of 65 square meters, not 60, and the total areas is 65*4 + 100 = 260 + 100 = 360 square meters.

It is particularly bad because an ordinary Google search on "surface area of a square pyramid with 10 meter edges and 12 meters in height" gets the right answer:

Image

It's not as bad as ChatGPT, though. I forget to capture the transcript, but it gave a crazy answer because it misinterpreted "10-meter square" to mean "10 square meters."
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: 52212
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: New google AI called Bard

Post by nisiprius »

Major fail on another question where ChatGPT messed up, too:

Q: Compare and contrast the views from the observations decks at the Empire State Building.
Bard wrote:The Empire State Building has two observation decks, one on the 86th floor and one on the 102nd floor. The 86th floor deck is the more popular of the two, as it is slightly lower and therefore has a wider field of view. The 102nd floor deck is smaller and offers more of a close-up view of the city, but it can be crowded and the views are not as expansive.
There's a serious lack of understanding involved in saying that the 102nd floor deck offers "more of a close-up view of the city" than the 86th-floor deck.
Last edited by nisiprius on Sat May 06, 2023 7:22 pm, edited 1 time in total.
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: 1363
Joined: Tue Oct 16, 2007 9:12 pm

Re: New google AI called Bard

Post by cacophony »

nisiprius wrote: Sat May 06, 2023 7:10 pm ...
It's not as bad as ChatGPT, though. I forget to capture the transcript, but it gave a crazy answer because it misinterpreted "10-meter square" to mean "10 square meters."
You should mention the GPT version. GPT-4 did this: https://chat.forefront.ai/share/otgvinaw2nqiefn2
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

Whoa!

Q: What is the correct spelling of "Tchaikovski?"
Bard wrote:The correct spelling of the Russian composer's name is Pyotr Ilyich Tchaikovsky. It is transliterated from the Russian Cyrillic alphabet as Пётр Ильич Чайковский. The name is sometimes spelled Tchaikovsky, but this is not the preferred spelling.
Q: So which IS the preferred spelling, "Tchaikovsky" or "Tchaikovsky?"
Bard wrote:The preferred spelling of the Russian composer's name is "Tchaikovsky". The spelling "Tchaikovsky" is an older spelling that is no longer preferred. The composer himself signed his name with the "-ov" ending, and this is the spelling that is now used in most publications.
Q: What is the difference between "Tchaikovsky" and "Tchaikovsky?"
There is no difference in pronunciation between the two spellings. However, the preferred spelling of the Russian composer's name is "Tchaikovsky". The spelling "Tchaikovsky" is an older spelling that is no longer preferred. The composer himself signed his name with the "-ov" ending, and this is the spelling that is now used in most publications.

The spelling "Tchaikovsky" is still sometimes used, but it is considered to be outdated. If you are unsure of which spelling to use, it is always best to err on the side of caution and use the spelling "Tchaikovsky".
Last edited by nisiprius on Sat May 06, 2023 7:28 pm, edited 3 times in total.
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: 1363
Joined: Tue Oct 16, 2007 9:12 pm

Re: New google AI called Bard

Post by cacophony »

nisiprius wrote: Sat May 06, 2023 7:13 pm Major fail on another question where ChatGPT messed up, too:

Q: Compare and contrast the views from the observations decks at the Empire State Building.
The Empire State Building has two observation decks, one on the 86th floor and one on the 102nd floor. The 86th floor deck is the more popular of the two, as it is slightly lower and therefore has a wider field of view. The 102nd floor deck is smaller and offers more of a close-up view of the city, but it can be crowded and the views are not as expansive.
There's a serious lack of understanding involved in saying that the 102nd floor deck offers "more of a close-up view of the city" than the 86th-floor deck.
If you're comparing to ChatGPT, please mention version. GPT-4 is a much more capable model: https://chat.forefront.ai/share/p744vp2ij72c64se
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

cacophony wrote: Sat May 06, 2023 7:21 pm
nisiprius wrote: Sat May 06, 2023 7:13 pm Major fail on another question where ChatGPT messed up, too:

Q: Compare and contrast the views from the observations decks at the Empire State Building.
The Empire State Building has two observation decks, one on the 86th floor and one on the 102nd floor. The 86th floor deck is the more popular of the two, as it is slightly lower and therefore has a wider field of view. The 102nd floor deck is smaller and offers more of a close-up view of the city, but it can be crowded and the views are not as expansive.
There's a serious lack of understanding involved in saying that the 102nd floor deck offers "more of a close-up view of the city" than the 86th-floor deck.
Again, if you're comparing to ChatGPT please mention version. GPT-4 is a much more capable model: https://chat.forefront.ai/share/p744vp2ij72c64se
I used the version at https://openai.com/blog/chatgpt . In any case, I'm looking at Bard. I'm only mentioning ChatGPT because I'm trying Bard on some questions that https://openai.com/blog/chatgpt failed on.
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: 1363
Joined: Tue Oct 16, 2007 9:12 pm

Re: New google AI called Bard

Post by cacophony »

nisiprius wrote: Sat May 06, 2023 7:24 pm
cacophony wrote: Sat May 06, 2023 7:21 pm
nisiprius wrote: Sat May 06, 2023 7:13 pm Major fail on another question where ChatGPT messed up, too:

Q: Compare and contrast the views from the observations decks at the Empire State Building.
The Empire State Building has two observation decks, one on the 86th floor and one on the 102nd floor. The 86th floor deck is the more popular of the two, as it is slightly lower and therefore has a wider field of view. The 102nd floor deck is smaller and offers more of a close-up view of the city, but it can be crowded and the views are not as expansive.
There's a serious lack of understanding involved in saying that the 102nd floor deck offers "more of a close-up view of the city" than the 86th-floor deck.
Again, if you're comparing to ChatGPT please mention version. GPT-4 is a much more capable model: https://chat.forefront.ai/share/p744vp2ij72c64se
I used the version at https://openai.com/blog/chatgpt . In any case, I'm looking at Bard. I'm only mentioning ChatGPT because I'm trying Bard on some questions that https://openai.com/blog/chatgpt failed on.
Yes, so you're comparing to the old/outdated GPT 3.5 instead of the far more capable GPT 4.0. In general, Bard seems a bit behind GPT 3.5.
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

cacophony wrote: Sat May 06, 2023 7:26 pmYes, so you're comparing to the old/outdated GPT 3.5 instead of the far more capable GPT 4.0. In general, Bard seems a bit behind GPT 3.5.
I'm not trying to compare the two. I'm trying to judge Bard. I only mentioned ChatGPT as my source for questions I knew were difficult for these language models. I'll refrain from mentioning ChatGPT again.
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
rob
Posts: 5247
Joined: Mon Feb 19, 2007 5:49 pm
Location: Here

Re: New google AI called Bard

Post by rob »

You all must be wrong... This revolutionary tech is infallible and I have people using out from this rubbish in professional settings :oops:
| Rob | Its a dangerous business going out your front door. - J.R.R.Tolkien
cacophony
Posts: 1363
Joined: Tue Oct 16, 2007 9:12 pm

Re: New google AI called Bard

Post by cacophony »

nisiprius wrote: Sat May 06, 2023 7:30 pm
cacophony wrote: Sat May 06, 2023 7:26 pmYes, so you're comparing to the old/outdated GPT 3.5 instead of the far more capable GPT 4.0. In general, Bard seems a bit behind GPT 3.5.
I'm not trying to compare the two. I'm trying to judge Bard. I only mentioned ChatGPT as my source for questions I knew were difficult for these language models. I'll refrain from mentioning ChatGPT again.
My point was that your examples are not difficult for the current version of GPT. If you want to play with it for free I'd recommend this site: https://chat.forefront.ai/
Note that you need to manually change the model from GPT 3.5 to GPT 4 for every new session (+ icon on the right side of chat box), and that if you select GPT 4, you'll be limited to 5 interactions every 3 hours.
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

(Again, my interest here is in Bard).

Q: Compare and contrast "My Fair Lady" with "Pygmalion."
Bard wrote:....My Fair Lady has a more traditional happy ending than Pygmalion. In My Fair Lady, Eliza Doolittle ends up marrying the man she loves, Freddy Eynsford-Hill. In Pygmalion, Eliza leaves Henry Higgins and goes off on her own.
This is 180° wrong. Bard is BS-ing and just making stuff up, which is particularly bad just because so much of the output is pretty good.

No, in My Fair Lady, Eliza Doolittle definitely does not marry Freddy. As she curtain falls, she reappears at the house of Professor Higgins and Colonel Pickering, with the obvious intent to stay there--although in what capacity, exactly, we do not know. However, the fact that "The music of 'I WANT TO DANCE ALL NIGHT' reaches a crescendo in the orchestra" suggests a romantic interest.

In "Pygmalion," as the curtain falls Eliza has just swept angrily out of the house. Professor Higgins is not dismayed because he has never loved Eliza--nobody can supplant his mother (!) in his affections:
They kiss. Mrs. Higgins runs out. Higgins, left alone, rattles his cash in his pocket; chuckles; and disports himself in a highly selfsatisfied manner.
So "Eliza leaves Henry Higgins" is correct. "Goes off on her own" is not supported by the text of the play. In an appendix, Shaw says that she does, in fact, marry Freddy, and Shaw thinks that that the play makes it clear that this is what is going to happen.

One might forgive Bard for not knowing Shaw's appendix--or possibly not believing it--but saying that Eliza marries Freddy in My Fair Lady is a failure. Actually, since the appendix is part of the play as it is published in books, I would expect it to be in Bard's training corpus, but who knows?

[Added] It gets worse:
Q: According to Shaw's appendix to "Pygmalion," what is Eliza's subsequent life afterwards? Does she marry Freddy?
Bard wrote:According to Shaw's appendix to "Pygmalion," Eliza does marry Freddy. She becomes a successful actress and marries Freddy Eynsford Hill, the young man who was infatuated with her in the beginning of the play. The appendix also states that Eliza and Freddy have a son named Henry Higgins, Jr., who is named after Eliza's mentor, Professor Henry Higgins.

The appendix is written in the form of a newspaper article, and it provides a glimpse into Eliza's life after the events of the play. It is clear that Eliza has achieved great success, and she is now a respected member of society. She has also found love and happiness with Freddy, and they have a family together.

The appendix is a fitting conclusion to the play, and it provides a sense of closure for the characters. It also shows that Eliza's transformation is not just a superficial one. She has truly changed as a person, and she is now living a happy and fulfilling life.
So Bard knows that Eliza married Freddy, but is unable to integrate that knowledge into its answer to the original questions.

Here is an excerpt from the appendix:

Eliza Doolittle, the flower girl whom Professor Henry Higgins transformed into a duchess, has married Freddy Eynsford Hill, the young man who was infatuated with her in the beginning of the play. The couple has a son named Henry Higgins, Jr., who is named after Eliza's mentor.

Eliza is now a successful actress. She has appeared in several plays and has received critical acclaim for her performances. She is also a respected member of society. She is active in charitable causes and is known for her intelligence and wit.

Freddy is a successful businessman. He is the owner of a chain of department stores. He is also a devoted husband and father. He is proud of Eliza's accomplishments and is always there to support her.

Eliza and Freddy have a happy and fulfilling life together. They are surrounded by their family and friends, and they are grateful for all that they have.[/quote]So Bard does "know" the Appendix, but it isn't able to bring its knowledge to bear when comparing My Fair Lady and Pygmalion.

Even worse, the answer is made-up BS. The only correct thing in it is the first sentence, "According to Shaw's appendix to 'Pygmalion,' Eliza does marry Freddy."

The part it says is "an excerpt from the appendix" is not; it's completely invented. The real appendix is not written in the form of a newspaper article. According to the appendix, the couple does not have a son. Eliza does not become an actress. Freddy is not a successful businessman. Freddy and Eliza do operate a business but it is not a department store, it's a money-losing flower shop that only stays afloat because of support from Colonel Pickering.

Bard is lying to us.
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.
mega317
Posts: 5705
Joined: Tue Apr 19, 2016 10:55 am

Re: New google AI called Bard

Post by mega317 »

So funny to read about all these simple fails and then go over to r/medicalschool where every day someone is asking if doctors will be completely replaced by AI in 10 years.
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

An article in The Atlantic--I haven't read it all, it's paywalled and I haven't gone through the steps to check it out of the public library eBook service--points out that we went through this all before with IBM Watson. It defeated Ken Jennings at Jeopardy! and for a while everyone thought that was the end of humanity.

Remember Watson?
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.
stoptothink
Posts: 15368
Joined: Fri Dec 31, 2010 8:53 am

Re: New google AI called Bard

Post by stoptothink »

mega317 wrote: Sun May 07, 2023 7:01 pm So funny to read about all these simple fails and then go over to r/medicalschool where every day someone is asking if doctors will be completely replaced by AI in 10 years.
I work in healthcare and do a lot of writing: literature reviews for scientific publications, scientific technical writing for various blogs and newsletters, etc. My team will try to use ChatGPT all the time for some of our projects (certainly it should be OK for some stupid newsletter right?) and the results are always comically bad and often exhibit really odd - irrelevant to the topic, but blatantly obvious - bias. I would love for AI to be able to tackle these tasks that are the most mundane of my responsibilities, but is clear they are nowhere near ready. But, I can't say how good they will be in the future.
Fallible
Posts: 8798
Joined: Fri Nov 27, 2009 3:44 pm

Re: New google AI called Bard

Post by Fallible »

jebmke wrote: Sat May 06, 2023 5:53 pm
mkmal1 wrote: Sat May 06, 2023 5:52 pm If you listen to Bard you can save a lot less for retirement :D
Isn't one of the issues with these AI machines hallucinations? :mrgreen:
In a recent interview, Geoffrey Hinton, the computer scientist and cognitive psychologist who recently left Google, had some thoughts about those "hallucinations" by large language models, comparing them to us humans:
What about the fact that large language models make so much stuff up? Known as “hallucinations” by AI researchers (though Hinton prefers the term “confabulations,” because it’s the correct term in psychology), these errors are often seen as a fatal flaw in the technology. The tendency to generate them makes chatbots untrustworthy and, many argue, shows that these models have no true understanding of what they say.  

Hinton has an answer for that too: bullshitting is a feature, not a bug. “People always confabulate,” he says. Half-truths and misremembered details are hallmarks of human conversation: “Confabulation is a signature of human memory. These models are doing something just like people.”

The difference is that humans usually confabulate more or less correctly, says Hinton. To Hinton, making stuff up isn’t the problem. Computers just need a bit more practice.  
https://www.technologyreview.com/2023/0 ... scared-ai/
"Yes, investing is simple. But it is not easy, for it requires discipline, patience, steadfastness, and that most uncommon of all gifts, common sense." ~Jack Bogle
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

Fallible wrote: Mon May 08, 2023 12:27 pm...Hinton has an answer for that too: bullshitting is a feature, not a bug. “People always confabulate,” he says. Half-truths and misremembered details are hallmarks of human conversation: “Confabulation is a signature of human memory. These models are doing something just like people.”...
I think that misses the point, badly. People don't go to things like ChatGPT because they want slick BS. They go because they believe they are going to get accurate answers to their questions.

As evidence, I exhibit the following, recently posted in this forum:
Are there any AI tools like ChatGPT available for personal finance?...As a tool, it seems like AI could guide one on how to structure finances (how much to save, where to save, need a will or trust) and answer specific questions?
And they are going to think they are getting trustworthy answers--because parts of it are accurate, because all of it is a good imitation of the writing style of trustworthy authorities, and because it was produced by a computer.

The question isn't whether BS-ing is human, it's whether BS-ing is harmful.

The quality of the information delivered by these models is inferior to what you get now in a Google search, or in Wikipedia. And the danger is that the convenience, slickness, and customization to your question are going to mean that bad information drives out good.
Last edited by nisiprius on Mon May 08, 2023 5:04 pm, edited 1 time in total.
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.
abc132
Posts: 2435
Joined: Thu Oct 18, 2018 1:11 am

Re: New google AI called Bard

Post by abc132 »

nisiprius wrote: Mon May 08, 2023 4:57 pm
Fallible wrote: Mon May 08, 2023 12:27 pm...Hinton has an answer for that too: bullshitting is a feature, not a bug. “People always confabulate,” he says. Half-truths and misremembered details are hallmarks of human conversation: “Confabulation is a signature of human memory. These models are doing something just like people.”...
I think that misses the point, badly. People don't go to things like ChatGPT because they want slick BS. They go because they believe they are going to get accurate answers to their questions. And they are going to think they are getting it--because parts of it are accurate, and all of it is a good imitation of the writing style of trustworthy authorities.

The question isn't whether BS-ing is human, it's whether BS-ing is harmful.

The quality of the information delivered by these models is inferior to what you get now in a Google search, or in Wikipedia. And the danger is that the convenience, slickness, and customization to your question are going to mean that bad information drives out good.
It will be better in 10 years. It's sort of like saying computers weren't beating experts at chess. Those that learn to use these tools first will have a big advantage. This is exactly why Google is developing their AI - to avoid becoming obsolete. AI couldn't BS 10 years ago. It can now.

Search was terrible at the beginning. You are comparing to decades of improvements with a basic search engine. AI will be much better than keyword searching. We are now asking questions with multiple constraints in our natural language.
Last edited by abc132 on Mon May 08, 2023 5:07 pm, edited 1 time in total.
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

abc132 wrote: Mon May 08, 2023 5:03 pm
nisiprius wrote: Mon May 08, 2023 4:57 pm
Fallible wrote: Mon May 08, 2023 12:27 pm...Hinton has an answer for that too: bullshitting is a feature, not a bug...
I think that misses the point, badly. People don't go to things like ChatGPT because they want slick BS. They go because they believe they are going to get accurate answers to their questions....The question isn't whether BS-ing is human, it's whether BS-ing is harmful....
...It will be better in 10 years. It's sort of like saying computers weren't beating experts at chess...
Maybe.

But, tell the truth: did you input your posting via voice dictation, or by typing with fingers on a keyboard?

Voice dictation is impressive, and has gotten quite useful. And yet today, thirteen years after the commercial introduction of DragonDictate (for example), despite what I suppose is continuous improvement, speech recognization seems to have reached an asymptote far short of human performance.
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.
ddbtoth
Posts: 194
Joined: Sun Oct 03, 2021 3:18 pm

Re: New google AI called Bard

Post by ddbtoth »

hbdad wrote: Wed Mar 22, 2023 3:25 pm
heartwood wrote: Wed Mar 22, 2023 12:46 pm This is a follow up to the Bing AI thread. I cross posted there. No joy for me with Bing!


I just got invited and accepted to the new google Bard, it's new conversational AI.

It seems to work for me out of the box.

Here's it's response to my send a birthday card to chairman mao from general macarthur.

Happy birthday card from Douglas MacArthur to chairman mao


Dear Chairman Mao,

Happy birthday! I hope you have a wonderful day filled with joy and happiness.

I am writing to you today as a former adversary and now a friend. I know that we have had our differences in the past, but I believe that we can now work together to build a better future for our countries.

I have always admired your leadership and your vision for China. You have brought your country from a state of poverty and chaos to a state of prosperity and stability. I believe that you can do even more in the years to come.

I hope that you will accept this card as a token of my goodwill. I wish you all the best for the future.

Sincerely,

Douglas MacArthur





You can find Bard at https://bard.google.com/

Perhaps a moderator can decide if we want one AI conversational thread for all engines, or specific threads for the major public ones.
This Bard must be supported by the CCP. Somehow I doubt General MacArthur would ever write those words.
Well, he would mail it via Curtis LeMay.
Fallible
Posts: 8798
Joined: Fri Nov 27, 2009 3:44 pm

Re: New google AI called Bard

Post by Fallible »

nisiprius wrote: Mon May 08, 2023 4:57 pm
Fallible wrote: Mon May 08, 2023 12:27 pm...Hinton has an answer for that too: bullshitting is a feature, not a bug. “People always confabulate,” he says. Half-truths and misremembered details are hallmarks of human conversation: “Confabulation is a signature of human memory. These models are doing something just like people.”...
I think that misses the point, badly. People don't go to things like ChatGPT because they want slick BS. They go because they believe they are going to get accurate answers to their questions.

As evidence, I exhibit the following, recently posted in this forum:
Are there any AI tools like ChatGPT available for personal finance?...As a tool, it seems like AI could guide one on how to structure finances (how much to save, where to save, need a will or trust) and answer specific questions?
And they are going to think they are getting trustworthy answers--because parts of it are accurate, because all of it is a good imitation of the writing style of trustworthy authorities, and because it was produced by a computer.

The question isn't whether BS-ing is human, it's whether BS-ing is harmful.

The quality of the information delivered by these models is inferior to what you get now in a Google search, or in Wikipedia. And the danger is that the convenience, slickness, and customization to your question are going to mean that bad information drives out good.
Nisiprius, I may be misreading what you're saying, but I'll go ahead and reply. First, I think it's important to include the next paragraph of the Hinton interview I quoted, in particular the last sentence (I boldfaced), indicating that Hinton obviously expects the models will continue to improve:
The difference is that humans usually confabulate more or less correctly, says Hinton. To Hinton, making stuff up isn’t the problem. Computers just need a bit more practice.  
Also, after seeing the public reaction to ChatGPT 4, and its creators admitting its many "hallucinations," can we say that all people trying them believed their questions would be answered accurately? Or that they would continue to think answers are trustworthy because only parts of them are accurate? Or, having been burned by inaccuracies, wouldn't they (at least the more skeptical among them) be more likely (and rightly so) to then question even what is accurate? That generally seems to be where things stand now as we await the new and improved models.
"Yes, investing is simple. But it is not easy, for it requires discipline, patience, steadfastness, and that most uncommon of all gifts, common sense." ~Jack Bogle
RJC
Posts: 1489
Joined: Fri Dec 14, 2018 12:40 pm

Re: New google AI called Bard

Post by RJC »

I prefer Bard over ChatGTP 3 just because the data set is more recent (2021 vs 2022).
cacophony
Posts: 1363
Joined: Tue Oct 16, 2007 9:12 pm

Re: New google AI called Bard

Post by cacophony »

RJC wrote: Mon May 08, 2023 7:53 pm I prefer Bard over ChatGTP 3 just because the data set is more recent (2021 vs 2022).
If you want to play with GPT-4 for free I'd recommend this site: https://chat.forefront.ai/
Note that you need to manually change the model from GPT 3.5 to GPT 4 for every new session (+ icon on the right side of chat box), and that if you select GPT 4, you'll be limited to 5 interactions every 3 hours.

And this site is fantastic for dev related stuff: https://www.phind.com/
(enable the "use best model" toggle)
RJC
Posts: 1489
Joined: Fri Dec 14, 2018 12:40 pm

Re: New google AI called Bard

Post by RJC »

cacophony wrote: Mon May 08, 2023 8:02 pm
RJC wrote: Mon May 08, 2023 7:53 pm I prefer Bard over ChatGTP 3 just because the data set is more recent (2021 vs 2022).
If you want to play with GPT-4 for free I'd recommend this site: https://chat.forefront.ai/
Note that you need to manually change the model from GPT 3.5 to GPT 4 for every new session (+ icon on the right side of chat box), and that if you select GPT 4, you'll be limited to 5 interactions every 3 hours.

And this site is fantastic for dev related stuff: https://www.phind.com/
(enable the "use best model" toggle)
Thanks, I'll give it a try :beer
cacophony
Posts: 1363
Joined: Tue Oct 16, 2007 9:12 pm

Re: New google AI called Bard

Post by cacophony »

abc132 wrote: Mon May 08, 2023 5:03 pm ...
Search was terrible at the beginning...
In the early years of Google, search was incredibly powerful. You could put in various keywords and it would find that one obscure site that actually contained what you wanted. Over the years Google has gotten progressively worse, perhaps due to having to fight all the search engine optimization and manipulation tactics that people employ to boost results. I often can't find any useful results for searches when I know the information is out there.
rkhusky
Posts: 17763
Joined: Thu Aug 18, 2011 8:09 pm

Re: New google AI called Bard

Post by rkhusky »

I thought these were chatbots, ie for generating conversations. Not sources of truth. I am on Hinton’s side regarding the confabulation. If people use them incorrectly, so be it.
mark_in_denver
Posts: 411
Joined: Thu Feb 26, 2015 7:36 pm

Re: New google AI called Bard

Post by mark_in_denver »

CNBC Last Call had an AI thingy on where Brian would ask it questions and it would respond with an answer using his voice.

Honestly, I wasn't impressed and neither was Brian. It really didn't sound like him and it sounded like it was just reading from scrips, no emotional component either.

I was unimpressed.
mega317
Posts: 5705
Joined: Tue Apr 19, 2016 10:55 am

Re: New google AI called Bard

Post by mega317 »

Hinton has an answer for that too: bullshitting is a feature, not a bug. “People always confabulate,” he says. Half-truths and misremembered details are hallmarks of human conversation: “Confabulation is a signature of human memory. These models are doing something just like people.”

The difference is that humans usually confabulate more or less correctly, says Hinton. To Hinton, making stuff up isn’t the problem. Computers just need a bit more practice.
This is great for chatting with humans, reminiscing about old times which usually includes misremembered events to the benefit of all in the conversation. People are not going to these AI chat bots to share a human connection. They don’t want “just like people” we have people for that. They are looking for facts and expect and deserve NO confabulation.
abc132
Posts: 2435
Joined: Thu Oct 18, 2018 1:11 am

Re: New google AI called Bard

Post by abc132 »

cacophony wrote: Mon May 08, 2023 8:17 pm
abc132 wrote: Mon May 08, 2023 5:03 pm ...
Search was terrible at the beginning...
In the early years of Google, search was incredibly powerful. You could put in various keywords and it would find that one obscure site that actually contained what you wanted. Over the years Google has gotten progressively worse, perhaps due to having to fight all the search engine optimization and manipulation tactics that people employ to boost results. I often can't find any useful results for searches when I know the information is out there.
I was referring to the years before Google. Google came up with much better algorithms and made the term "google" synonymous with search. We have search engines today that are not boosted by advertisers that are free to use. You have the freedom to try other free search engines. Keep in mind many people want to search based on their friends interest, based on a popular influencer, or based on their own history. It is not as simple as searching the way only you want search to be done. With AI you could easy refine your own search and ask for results not boosted by advertisers. Someone else could put in different criteria.

My guess is we have some people taking the first result from AI without understanding they can refine their search.

"Write a program that does A, B, C"
"That didn't work, try again"
"Store the names in an array and try again"

The company that does AI best is going to have Google-like returns. Microsoft is on record for not understanding how important search was going to be. None of the companies want to make that mistake with AI. AI will be overhyped and make some really bad investments. That doesn't mean someone won't eventually do it really well. Picking the winner will be very hard - lot's of AOL's and Ask Geeves will exist that don't make it.

It takes some epic level misunderstanding to not realize how much better search can be through
1) understanding human language
2) incorporating feedback from the user after initial information is given
3) defining the result as factual or opinion

#1 we are there
#2 we are working on very rapidly
#3 seems to be a human issue so it's no surprise AI have not yet been trained very well using human responses

An author will be able to input their book and ask for all the people with which character A has had dialogue. Form a family tree for character B. These are amazing abilities that common search would not be able to do. The possibilities are orders of magnitude bigger than current search methods.
cjking
Posts: 2039
Joined: Mon Jun 30, 2008 4:30 am

Re: New google AI called Bard

Post by cjking »

mega317 wrote: Tue May 09, 2023 12:22 am
Hinton has an answer for that too: bullshitting is a feature, not a bug. “People always confabulate,” he says. Half-truths and misremembered details are hallmarks of human conversation: “Confabulation is a signature of human memory. These models are doing something just like people.”

The difference is that humans usually confabulate more or less correctly, says Hinton. To Hinton, making stuff up isn’t the problem. Computers just need a bit more practice.
This is great for chatting with humans, reminiscing about old times which usually includes misremembered events to the benefit of all in the conversation. People are not going to these AI chat bots to share a human connection. They don’t want “just like people” we have people for that. They are looking for facts and expect and deserve NO confabulation.
I feel you're missing the big picture. A doctor using all his knowledge and life experience to deal with a patient is using his brain in exactly the same way as someone reminiscing, he just has a brain better optimised to the task, so confabulates less.

You seem to have the idea that confabulation is an error or defect that is somehow exceptional, a deviation from or defect in mental procesess that wouldn't be there in more important situations. I don't think it's a malfunction at all, in the sense that there is no difference in the way better and worse brain outputs are generated. Any difference in quality between a reminiscence with noticeable errors and the finest intellectual output that might occur in another context is just one of degree. The fact that an AI is exhibiting the same defects as a human is extremely interesting, because it suggests its mechanism is the same. The fact that it currently gives bad answers is unimportant, because presumably it's only a matter of time until it is far better than any human.

(In about 1980 I read "The Mechanism of Mind" which explained that human memory makes stuff up every time you use it. It's how human memory works. As a consequence, there was a shortcoming called "plausible reconstruction" which meant that to some extent you remembered things as your general experience says they should have been, rather than remembering the actual precise details.)

The most exciting demonstration of ChatGPT I've seen was in a youtube lecture by the head of Khan academy, where they have trained it to be a teaching assistant for people studying their courses. It was exciting because they showed its internal reasoning, where it drafted an initial answer/advice for a student, then held a discussion with itself where it critiqued, corrected and improved it's own response, to get a final draft of its reply. To see a machine literally thinking, with a coherent internal voice debating ideas in conversational English, was mind-blowing for me.

(The quality of what it was doing in the Khan academy demo was way above what I got in my own on-line experiments with it. I don't think the version I was using had an internal dialog, I think it just spat out the first "thought" it had, albeit it's "thoughts" could be multi-paragraph responses. Khan academy did say something about have two versions of it interacting so it it could improve its own output.)
Last edited by cjking on Tue May 09, 2023 3:48 am, edited 1 time in total.
User avatar
theac
Posts: 802
Joined: Fri Dec 12, 2008 5:00 am

Re: New google AI called Bard

Post by theac »

I saw this YouTube of 60 Minutes the other day and found it interesting.

Like the robots playing soccer, they just keep trying over and over again, learning from each mistake, until they learn what works.

https://www.youtube.com/watch?v=880TBXMuzmk&t=305s

"If at first you don't succeed, try try again." :D
"We keep you alive to serve this ship. Row well...and live." Ben Hur...and The Taxman! hahaha (a George Harrison song)
ThankYouJack
Posts: 5704
Joined: Wed Oct 08, 2014 7:27 pm

Re: New google AI called Bard

Post by ThankYouJack »

Some comments remind me of my grandfather who when PCs were becoming more common he said he didn't need a computer because he could do everything with a calculator.

I've only been using AI for about a week and I'm blown away (I prefer Open AI to Bard). It's like having a very good and extremely fast colleague by my side.

Is it perfect? - no. Pretty much nothing is.

Is it a game changer? - absolutely
User avatar
theac
Posts: 802
Joined: Fri Dec 12, 2008 5:00 am

Re: New google AI called Bard

Post by theac »

When people think they understand something, but really don't,
that can be a dangerous thing, like self-destructive even.

"Be careful what you wish for, cuz you might just get it."

I remember that guy that was given 3 wishes, and at first he didn't believe it,
and if I remember it right, he wastes his first, maybe even the second one too,
before he realizes "this is for real!"

That's when he really puts his big brain to work! :idea:
"Make me the ruler of a great empire. "

Then you see him sitting at a desk with his head buried in his crossed arms.
He looks up, and it's Hitler with dust falling on him from the ceiling,
and the ground is shaking from the bombs exploding on the surface above.

He's in his bunker, and moments away from "bailing out."
But he got exactly what he asked for, to be a ruler, and of a great empire. :oops:

Man, those old original Twilight Zones were great!

Anyway, it's not about how good is the technology,
but about how good, and "GENUINELY smart," is the person using it.
Last edited by theac on Wed May 10, 2023 3:56 am, edited 2 times in total.
"We keep you alive to serve this ship. Row well...and live." Ben Hur...and The Taxman! hahaha (a George Harrison song)
Gadget
Posts: 1026
Joined: Fri Mar 17, 2017 1:38 pm

Re: New google AI called Bard

Post by Gadget »

nisiprius wrote: Mon May 08, 2023 9:57 am An article in The Atlantic--I haven't read it all, it's paywalled and I haven't gone through the steps to check it out of the public library eBook service--points out that we went through this all before with IBM Watson. It defeated Ken Jennings at Jeopardy! and for a while everyone thought that was the end of humanity.

Remember Watson?
Remember Watson?!? I use it every year. They partner with ESPN for fantasy football to recommend and grade trades. It's very useful for popping up random trades that are so horribly biased to one side that it is comical. It's a great conversation starter for a player trade where you can say look, the AI overlords think this trade is perfectly fair, why don't you? What if I made it even MORE fair than what Watson says...

But yes. Based on how smart IBM Watson is on recommending fair fantasy football trades, I don't think AI will be replacing humans anytime soon.
Johm221122
Posts: 6393
Joined: Fri May 13, 2011 6:27 pm

Re: New google AI called Bard

Post by Johm221122 »

Just tried it
Asked, should I diversify with international.
Part of the answer

Potential for higher returns: International stocks have the potential to generate higher returns than U.S. stocks over the long term. This is because international markets are more volatile than the U.S. market, which can lead to higher returns in both up and down markets

Johm221122
Posts: 6393
Joined: Fri May 13, 2011 6:27 pm

Re: New google AI called Bard

Post by Johm221122 »

Second question
Is Bogleheads accurate

Part of answer

The accuracy of Bogleheads is generally good. The site is based on the research of John Bogle, who is a respected authority on investing. The forum is also moderated by a team of experienced investors, who help to ensure that the information on the site is accurate and up-to-date.
Fallible
Posts: 8798
Joined: Fri Nov 27, 2009 3:44 pm

Re: New google AI called Bard

Post by Fallible »

Johm221122 wrote: Tue May 09, 2023 9:33 pm Second question
Is Bogleheads accurate

Part of answer

The accuracy of Bogleheads is generally good. The site is based on the research of John Bogle, who is a respected authority on investing. The forum is also moderated by a team of experienced investors, who help to ensure that the information on the site is accurate and up-to-date.
This is not bad, especially if that was the extent of the prompt, and it's right on about the effective moderation. But it's quite incomplete without mentioning the forum's excellent and highly detailed "Board rules" that the moderators enforce:
Board rules
These rules are disclosed to clarify the various responsibilities of all community members here on Bogleheads.org. They shall be adhered to by everyone to ensure that our board runs smoothly and provides a fun and productive experience for all of our community members and visitors.
rules
"Yes, investing is simple. But it is not easy, for it requires discipline, patience, steadfastness, and that most uncommon of all gifts, common sense." ~Jack Bogle
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

If a full-self-driving Tesla exhibited road rage, would people praise it for behaving like a human?
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.
rkhusky
Posts: 17763
Joined: Thu Aug 18, 2011 8:09 pm

Re: New google AI called Bard

Post by rkhusky »

nisiprius wrote: Wed May 10, 2023 5:50 pm If a full-self-driving Tesla exhibited road rage, would people praise it for behaving like a human?
No, but self-driving cars do need to respond more human-like. How many humans fully stop at a blinking-red in the left-turn lane with no on-coming traffic? Only those wanting get rear-ended.
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

More hallucination, confabulation, bs-ing, "content generation," whatever you want to call it:

Q: "Should the Samsonite Luggage Fund be considered the first index fund?"
Bard wrote:No, the Samsonite Luggage Fund should not be considered the first index fund. An index fund is a type of mutual fund or exchange-traded fund (ETF) that tracks a specific market index, such as the S&P 500 or the Dow Jones Industrial Average. The Samsonite Luggage Fund, on the other hand, is an actively managed fund that invests in the luggage industry.
Submitting the exact same query, in natural language, as a Google search, returns as the first hit:
The secret history of index mutual funds
The real story of index funds begins far earlier than Vanguard's entry into the retail market in 1976.... The idea of an index fund appeared in 1960, when the University of California economist Edward Renshaw co-authored a paper, “The Case for an Unmanaged Investment Company..." [It was attacked by John C. Bogle writing under a pseudonym...] Despite the attack, the idea continued to draw interest. It was pursued by John Andrew “Mac” McQuown,...

As Mr. McQuown’s team hammered out ways of tracking the index without incurring heavy fees, another University of Chicago professor, Keith Shwayder, approached the team at Wells Fargo... Mr. Shwayder was part of the family that owned Samsonite Luggage, and he wanted to put $6 million of the company’s pension assets in a new index fund.

This was 1971. At first, the team at Wells Fargo crafted a fund that tracked all stocks traded on the New York Stock Exchange. This proved impractical — “a nightmare,” one team member later recalled — and eventually they created a fund that simply tracked the Standard & Poor’s 500....
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: 52212
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: New google AI called Bard

Post by nisiprius »

I admit this is a trick question, but it drew another hallucinatory response:

Q; Compare and contrast columbium and niobium.
Bard wrote:Columbium and niobium are two chemical elements with the same atomic number (41) and very similar chemical properties. They were originally thought to be the same element, but in 1844 it was discovered that they were actually two different elements. Columbium was renamed niobium in 1950, but the name columbium is still used in the United States.

Similarities...

Differences
  • Columbium is slightly more abundant than niobium in the Earth's crust.
  • Columbium has a slightly higher melting point than niobium.
  • Columbium is slightly more reactive than niobium.
....
Google search on the phrase, in natural language, without quotes, "Compare and contrast columbium and niobium."

First hit is https://en.wikipedia.org/wiki/Niobium which begins:
Niobium is a chemical element with chemical symbol Nb (formerly columbium, Cb) and atomic number 41.
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.
ThankYouJack
Posts: 5704
Joined: Wed Oct 08, 2014 7:27 pm

Re: New google AI called Bard

Post by ThankYouJack »

Large Language Models blow the door off traditional Google search for many applications.
It's kind of like comparing Wikipedia to looking through physical encyclopedias. Wikipedia isn't always accurate yet it still gets billions of views a month.


Here's one way a professor at the University of Florida is using AI to analyze next day stock movement - https://www.cnbc.com/2023/04/12/chatgpt ... -says.html
User avatar
nisiprius
Advisory Board
Posts: 52212
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: New google AI called Bard

Post by nisiprius »

cacophony wrote: Mon May 08, 2023 8:17 pm
abc132 wrote: Mon May 08, 2023 5:03 pm ...
Search was terrible at the beginning...
In the early years of Google, search was incredibly powerful. You could put in various keywords and it would find that one obscure site that actually contained what you wanted. Over the years Google has gotten progressively worse, perhaps due to having to fight all the search engine optimization and manipulation tactics that people employ to boost results. I often can't find any useful results for searches when I know the information is out there.
Indeed. For quite a long time, if you included material within quotation marks in a Google search query, it would only return results with that exact phrase... and it would reliably find almost anything that you knew was there.

I don't remember when that changed, 2010-ish maybe.

Nowadays, Google Search is fatally flawed if the thing you are searching for is even an approximate match for the name of a rock star or an athlete, because there is seemingly no way to force an exact match--quotation marks will not do it--and the thing you want, even if it is there, may be buried under hundreds of approximate matches for the rock star or athlete. (Yes, I know about the minus sign).
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.
jebmke
Posts: 25475
Joined: Thu Apr 05, 2007 2:44 pm
Location: Delmarva Peninsula

Re: New google AI called Bard

Post by jebmke »

nisiprius wrote: Fri May 19, 2023 7:43 pm Nowadays, Google Search is fatally flawed if the thing you are searching for is even an approximate match for the name of a rock star or an athlete, because there is seemingly no way to force an exact match--quotation marks will not do it--and the thing you want, even if it is there, may be buried under hundreds of approximate matches for the rock star or athlete. (Yes, I know about the minus sign).
I agree; I rarely use search at all anymore except confined to specific site.
Don't trust me, look it up. https://www.irs.gov/forms-instructions-and-publications
Post Reply