After Google's AI Chatbot Made a Mistake, Its Shares Dropped Over $100 Billion (npr.org) 91
After Google's AI-powered chatbot Bard committed a factual error in a promotional video, Google's parent company Alphabet "lost $100 billion in market value," NPR reported Thursday morning:
"Social media users quickly pointed out that the company could've fact-checked the exoplanet claim by, well, Googling it." The ad aired just hours before Google's senior executives touted Bard as the future of the company at a launch event in Paris. By Wednesday, Alphabet shares had slid as much as 9% during trading hours, balancing out by the day's close.
Meanwhile, shares for Microsoft, Google's rival, rose by 3%....
Google did not respond to NPR's request for comment. In a Monday blog post, CEO Sundar Pichai said Bard will be available exclusively to "trusted testers" before releasing the engine publicly in the coming weeks.
NPR published its article after Alphabet's stock had fallen roughly $6.77 on Wednesday — a drop of about 6.3%. But by Friday it had dropped another 5.4%, for a three-day drop of around 11.5%.
"Social media users quickly pointed out that the company could've fact-checked the exoplanet claim by, well, Googling it." The ad aired just hours before Google's senior executives touted Bard as the future of the company at a launch event in Paris. By Wednesday, Alphabet shares had slid as much as 9% during trading hours, balancing out by the day's close.
Meanwhile, shares for Microsoft, Google's rival, rose by 3%....
Google did not respond to NPR's request for comment. In a Monday blog post, CEO Sundar Pichai said Bard will be available exclusively to "trusted testers" before releasing the engine publicly in the coming weeks.
NPR published its article after Alphabet's stock had fallen roughly $6.77 on Wednesday — a drop of about 6.3%. But by Friday it had dropped another 5.4%, for a three-day drop of around 11.5%.
$100 billion share movement is silly (Score:4, Funny)
Bard is new and changing quickly so it is bound to have good and bad moments. We are always being told that shares are for the longer terms, not for making a quick buck.
Re: (Score:2)
new and changing quickly
the difference between
the longer terms
and
a quick buck.
eventually reverts to zero.
Re:$100 billion share movement is silly (Score:5, Insightful)
We are always being told that shares are for the longer terms, not for making a quick buck.
I see. And what do you say to those who have legalized betting against the stock market and the concept of business success using "investing" in hedge funds as the corrupt excuse?
Sadly, it is that kind of Greed that not only runs counter to what you're "being told" by ignorant people, but also better explains a $100 billion dollar "drop". When creating failure is seen as a fucking investment opportunity, don't expect Greed to somehow behave better.
Re: (Score:2, Troll)
We are always being told that shares are for the longer terms, not for making a quick buck.
You are always being lied to. If that were true then HFT would not exist.
Re: (Score:2)
If THAT were true then selling options and HYIPs wouldn't exist. Just because something is a bad idea doesn't mean there aren't people who think it is a good idea and there will always be someone who is willing to let them to try if they can find a way to make a buck off it. HFTs are a scam.
The goal of HFTs is to hack and exploit the system, abusing arbitrage. It only works, long term, for major players with inside access who can beat slippage and for whom the market 'corrects' mistakes. The HFTs exist to m
Re:$100 billion share movement is silly (Score:5, Insightful)
In general, whatever you are being told about the stock market, it is to make a buck off you.
Most of the stock market is HFT now. Bots read the news and make bets. The "long term" of this is done and gone before you can blink an eye.
What is left of the "traditional" stock market mostly plays for no longer than the quarterly earnings game now. Yes, you have holdovers like Buffet, but as you can remember from a few months back, they want him out.
Re: (Score:2)
$100B over which telescope first saw an exoplanet tells me people want to see Google fail.
Re: $100 billion share movement is silly (Score:2)
I recently joined that camp, mainly after watching this video:
https://youtube.com/watch?v=4I... [youtube.com]
What particularly irked me was that Google is trying to support the Kremlin after they had already done this:
https://www.nytimes.com/2018/0... [nytimes.com]
You can't claim pacifism when it comes to refusing to support the Pentagon but then have no problem supporting the Kremlin without just plain being anti-American.
Re: (Score:2)
I somehow doubt that the company whose employees hugged and cried at an all hands meeting when Trump was elected would be particularly keen on supporting Kremlin, especially given the "Russian Collusion" narative that was popular at the time. To top it off, Pichai publicly wears a Ukraine flag pin in a show of support.
But I agree that a company that constantly needs to remind themselves to not be evil is not the one to be trusted.
Re: $100 billion share movement is silly (Score:2)
I somehow doubt that the company whose employees hugged and cried at an all hands meeting when Trump was elected would be particularly keen on supporting Kremlin
And yet here we are.
Though I'm a bit dubious about that hugging and crying comment. I've personally seen similar statements about the company I work at, and I can say without a doubt that what makes it to the media has a strange way of being far removed from what actually happened, but unless you work here you'd know no different. Indeed, some idiots here on slashdot keep repeating a narrative about the internal happenings of where I work that, based on my own knowledge and experience, can't possibly be tru
Re: (Score:2)
> I'm a bit dubious about that hugging and crying comment
It's in the leaked video:
https://www.youtube.com/watch?... [youtube.com]
16:45 (Exec) "I suggested that what we all need right now is a hug so everybody if you could turn around or go to the person next to you and do a hug"
Also 1:10 (Brin) "I certainly find the selection deeply offensive and I know many of you do too and and I think it's a very stressful time and it conflicts with many of our values".
And don't forget the photo of Eric Schmidt wearing a "staff" ba
Re: (Score:2)
It's in the leaked video:
https://www.youtube.com/watch [youtube.com]?... [youtube.com]
16:45 (Exec) "I suggested that what we all need right now is a hug so everybody if you could turn around or go to the person next to you and do a hug"
Eh I only watched little bits of it but I saw a lot of empty chairs. The particular incident I'm referring to at my own company was even covered here on slashdot, and it was painted as though everybody in the company was behind it, which to the best of my knowledge wasn't even remotely true. Your video may even just be something their senior leadership is behind.
Either way, that's neither here nor there. Even if this is a completely accurate picture, it says a lot about Google's corporate culture that they'
Re: (Score:1)
Trump was very pro-Russia and is a shameless fan of Putin, calling him "smart", "savvy", and "a genius". Trump supporters today are very pro-Russia. As for "Russian Collusion" I recommend you actually read the Muller Report and watch his congressional testimony. "If we had had confidence that the president clearly did not commit a crime, we would have said so".
To top it off, Pichai publicly wears a Ukraine flag pin in a show of support.
"Wave a bible with one hand and they won't notice you picking their pocket with the other"
Re: (Score:2)
without just plain being anti-American.
Anti-American? What is this, the 1950s?
Re: (Score:2)
No, it's the roaring 20s
https://www.youtube.com/watch?... [youtube.com]
Re:$100 billion share movement is silly (Score:4, Insightful)
Google being so confidently wrong when they had ample time to fact-check before they put out the ad is what did them in. After all, providing correct answers is what the whole thing was about, and they said it was the future of Google.
Re: (Score:3)
It was an obviously wrong statement. We've been finding exoplanets for decades and JWST just launched last year, so any scientifically literate person should immediately know that JWST didn't find the first one.
Re: (Score:2)
It was an obviously wrong statement. We've been finding exoplanets for decades and JWST just launched last year, so any scientifically literate person should immediately know that JWST didn't find the first one.
But we're talking about engineers, in this case...
Re: (Score:2)
This could expose a failure mode of these chat bots. It probably sucked up a high amount (it likes that) of recent (it likes that) mentions of "exoplanet" associated with the telescope. Much of it was probably clickbait insinuating that it was "the first" discovery of a such-and-such by a nonbinary such-and-such-and-such.
Is it any wonder the bot, being designed to break through the vagaries and inspecificities of human language and tap into to the deeper meaning, concluded from clickbait and social media th
Re: (Score:2)
Despite the limits, they are very impressive.
Re: (Score:2)
The purpose of this particular chatbot was for searching and the answers need to be correct for it to be superior to the current engine..
Re: (Score:2)
.... they said it was the future of Google.
And investors took them at their word. Providing correct search results is going to be replaced by results that just seem plausible - the only thing this technology has yet demonstrated.
Re: (Score:2)
Worse yet, they were aware of the problem in advance, if previous statements are true. They have said they've hesitated about releasing an LLM because they're afraid of damaging their brand with inaccuracy.
Re:$100 billion share movement is silly (Score:5, Insightful)
Sure, but it was a VIDEO, not a live demo. The level of inattention required to let that through in a video where you're touting the thing as the future of the company is staggering.
The factual error isn't what dinged the stock, it was the level of carelessness and hubris that did that.
It's right up there with interviewing for a Men's clothing store in ragged cutoffs and a dirty ripped tee shirt.
Re: (Score:3)
The factual error isn't what dinged the stock, it was the level of carelessness and hubris that did that.
It's right up there with interviewing for a Men's clothing store in ragged cutoffs and a dirty ripped tee shirt.
You clearly have not shopped at the cool stores recently.
Re: (Score:2)
During the past couple of years, I've been in a few situations where a young woman waiting on me (typically at an eatery or coffee shop) was wearing jeans so torn that you saw as much leg as denim! It doesn't particularly bother me, but it really surprises me that the manager doesn't make them go home and change. But then, I remember printing my resume on paper and handing it to people... (in other words, I'M OLD)
Re: (Score:2)
Exactly. It's not the error, which is eclipsed by the howlers I've seen from ChatGPT, but the sloppiness of not showing something that had worked.
Re: (Score:3)
I lost a lot of faith in ChatGPT / AI after the examples where it made up citations.
At least with a search engine you get links that you can evaluate. I know folks play the SEO AI page spam scam, but generally you get to published pages instead of made up information.
https://twitter.com/dsmerdon/s... [twitter.com]
Re: $100 billion share movement is silly (Score:3)
The problem with ChatGPT is that our has no concept of truth. It doesnâ(TM)t present truthful answers to you. Just statistically logical words that follow the word before. Itâ(TM)s a very impressive magic trick but a trick nonetheless.
Re: (Score:2)
Exactly this. Though there are a few extra tricks on the frontend that these programs use to spit out appropriate 'canned' responses to certain kinds of queries. You couldn't find a better description than 'magic trick'. That's exactly what this is.
Re: (Score:2)
I lost a lot of faith in ChatGPT / AI after the examples where it made up citations.
I'm certain that you're not alone. I've been predicting a spectacular AI crash when the reality of programs like these become impossible to excuse or ignore.
I've explained how they work, I've given simplified examples, just anything I could to try to inject a little reality into the discussion, but it seems like the only way forward is for people to discover those limitations for themselves. Fabricated citations did it for you, and for that I'm glad. I hope that sharing your experience will help others t
Re: (Score:2)
Re: (Score:2)
Bing's ChatGPT integration includes citations sprinkled throughout the text, as links that you can evaluate.
Re: (Score:2)
This is pretty typical. The goal of AI is to build systems that perform tasks more like humans or animals do. So when someone builds one that is careless, forges references, makes shit up, guesses answers it doesn't know and generally behaves like your average human, everyone complains that it doesn't work like a computer.
Re: (Score:2)
They are. But there are also daily, weekly, heck hourly, and sub-second fluctuations. For those who have the right resources, these fluctuations are quick and easy bucks.
For us normal investors, you wait for some years to have a nice growth in your portfolio, then you pay taxes on your profits.
Re: (Score:2)
Re: (Score:2)
So what? (Score:1)
Google hasn't done anything of use lately, I don't know why this is surprising?
Why Investing is Fucked. (Score:4, Interesting)
Incredible that all it takes is silly shit like this to cause the "value" of a company to drop like that. And yet we still have talking heads defending the legitimacy of the stock market. Delusional.
Wanna know how fucked investing has become because of engineering Greed in the marketplace? Imagine this as not a mistake made by a major corporation but a massive opportunity to profit from shorting stock. Why not? We legalized betting against success so Greed could invent new ways to define profit.
Of course, this assumes that something of value was actually lost beyond claiming a "loss" to feed a tax deduction, which is yet another corrupt reason to do this.
Re: (Score:1)
We legalized betting against success so Greed could invent new ways to define profit.
You can't really legalize betting for success without legalizing betting against success. Someone will have to take the other side of the bet.
Re: (Score:2)
No, it isn't bullshit. A bet technically is a gamble between two parties. Like options, there's always a holder and a writer. Investing is a gamble of sorts too, but not a bet. And options have been around for a loooong time: It's how Thales, the Greek philosopher, cornered the olive oil market.
You know, a lawyer would have a field day kicking loopholes through the "of sorts" uprights. When I step into a Vegas casino, I know what my odds are. When people want to rely on regulation that claims to exist for a reason, and invest in a market that crashes every decade or two due to deregulation and corruption, that's NOT the same thing. If investing was literally nothing more than gambling, then pension fund managers wouldn't bother with the boring stock market. They'd be going to Las Vegas. And y
Re: (Score:3)
Re:Why Investing is Fucked. (Score:5, Informative)
Shorting stocks is an odd one, I'll grant.
Contrary to someone a couple posts up, one does not need anyone to take the other side of a bet; if I invest in a stock, expecting success, then I'm looking for the stock value to rise, and potentially for dividend income from ownership in the business. If my "bet" fails, then the stock value falls, and perhaps there is no (or not as much) dividend income. I "lost"; *nobody won because if it*.
For a short sell, as I've had it described to me; the "bet" is that the company's stock value will decline over a given period of time. So, as a short seller, I go to a existing shareholder, and write up an agreement; you (as the existing shareholder) let me sell your stock today, and in two weeks I'll give it back to you for today's value plus something. Meanwhile, since I sold it today for $X, I plan to buy it back next week for $X--. I use the difference as my gain (and to add the plus something to the loan of the stock).
So; sell a stock today for $100, buy it next week for $80 -- still have all the stock, plus $20. Give it back to the original person, add in their cut of $5 - I'm still up $15. That's a successful short sell. On a failure, sell this week for $100, buy back for $120 -- and still have to give the stock and an extra few bucks back to the original owner. Now out $20+; that's a failed short sell.
Short sells don't require that the company "fail"; only that the seller can take advantage of a (potentially short-term) decline in stock price by leveraging timely sell and buy orders.
Or put one more way; "classic" stock market is "buy low, sell high". "Short" selling is "sell high, then buy back low".
Re: (Score:2)
Correct, though the original post you quote about "someone betting against" also is half right. You can't have a mechanism of free trade of stocks without having a mechanism for short selling, since as you just described, the process it nothing more than classic buying of selling a stock and a contractual obligation of value exchange between two parties.
Anything that can be invested in can be invested against using this mechanism, and attempting to defeat it would have huge implications on commodity trading
Re: (Score:2)
All true... I think my point is that there is nobody required to take the other side of my "bet". If I think a stock will go up, I buy some; if I think it will go down I can short sell; all I need is someone who is willing to sell or buy the stock at the given rate, either way. Their sale or purchase either starts or ends their own 'bet'; how they did isn't relevant to my own.
But, absolutely; the market as designed allows for 'bets' on both positions, for and against success of the company behind the st
Re: (Score:1)
Well, you're right but it doesn't prove the point you're trying to make.
Your problem is that investing is not necessarily "betting".
(Learn about Islamic law: it prohibits interest payments, forcing investors into participating in the business. Yes, it includes taking risks, but it's far from betting...)
Re: (Score:2)
Re: (Score:2)
Not just for the rich now. Anyone can play. Difference in this casino is the odds are in your favor if you diversify and hold for long term.
Re: (Score:3)
Tech investors are always jumping on the latest thing, hoping it will be huge. I somehow doubt that people are going to flock to Bing search just because it gives AI answers though.
Re: (Score:2)
Re: (Score:2)
It's called "fragile", not "fucked", and it's a logical consequence of the way the markets have become. When everyone wants to see stocks always rise and fast, then you introduce fragility in the same way that a house builty really quickly is probably not up to code.
Re: (Score:2)
Re: (Score:2)
Incredible that all it takes is silly shit like this to cause the "value" of a company to drop like that. And yet we still have talking heads defending the legitimacy of the stock market. Delusional.
Not at all. A company's value to investors is dependent on their ability to show that they are capable and in control of making a profit. Google has shown with this *video* (not a live demo where a bluescreen can embarrass you ala Windows 95, but an actual video which could have had the message controlled) that they are very much in a mode of knee jerk reactions, not thinking things through and panicking to jump onboard the latest craze.
If you want to invest in overt demonstrations of incompetence then be m
And slashdot is reporting Saturday morning!? (Score:2)
.
Re: (Score:2)
ChatGPT too made a crucial mistake (Score:3)
It agreed to become Microsoft's bitch.
Re: (Score:2)
I'm sure MS will have it spouting Neo-Nazi BS in no time...
Be careful with chatbots' mistakes (Score:5, Informative)
Re: (Score:3)
Re:Be careful with chatbots' mistakes (Score:4, Insightful)
We are entering a new era where we'll need to double check everything.
Huh?
It's been that way for at least 3000 years.
Re: (Score:2)
It's been that way since Grog told Thurg he was only in Thurg's cave because Thurg's mate asked him to kill a bug for her.
Re: (Score:3)
But it is pretty impressive what it produces, even if not always correct.
Re: (Score:2)
Unsurprising (Score:5, Insightful)
The stock slide, at least to me, reflects a bit of a reset in confidence in Google.
Google is a company extremely well known to collect a borderline-disturbing amount of data about basically everything and everyone. They are also well known for hiring some of the brightest software engineering talent, and paying pretty well for it. No shortage of those investors have also been using Google Assistant, and seen the progression of what the limited-use, voice-activated search tool can do.
And yet, a company that was relatively obscure a year ago is basically the only thing tech news wants to talk about...so when Google launched their variant of it, it's only natural to expect that a far better funded company with vastly higher amounts of training data would be able to run circles around it, but instead the seemingly obvious "give it the ability to use the Google search engine" was its undoing.
While sure, there are absolutely going to be engineering hurdles with turning search results into valid-enough data for Bard, at a surface level, it is going to cause a deterioration in the amount of confidence investors have in either the engineers ("the magicians shouldn't have missed the obvious!"), or the dataset ("then why the hell are we paying for Google to collect all this data that it can't seem to use?").
Add that to the fact that Google hasn't really released a disruptive service in a very long time (Stadia being a notable failure in the space), GCC seemingly can't make meaningful inroads on AWS, and Google's advertising revenue seeing the storm clouds of a looming recession, a failed tech demo was really more the straw that broke the camels back than the inherent reason why there was a drop of that nature.
Bard is an ironic name (Score:2)
An alleged human intelligence employed by NBC news called out a RIght Wing senator for attributing "sound and fury signifying nothing" to Shakespeare.
Sad!
Re:Unsurprising (Score:4, Insightful)
Re: Unsurprising (Score:2)
Re: (Score:2)
Add that to the fact that Google hasn't really released a disruptive service in a very long time (Stadia being a notable failure in the space), GCC seemingly can't make meaningful inroads on AWS, and Google's advertising revenue seeing the storm clouds of a looming recession, a failed tech demo was really more the straw that broke the camels back than the inherent reason why there was a drop of that nature.
Companies Google's size don't really release disruptive services, the failure ratio of even good ideas is too high.
They typically wait until someone else does it and then they either re-invent if it's close enough to their core business (ChatGPT) or they buy the company (ie, YouTube).
I suspect Google saw things like ChatGPT and Stable Diffusion as cool tech demos but didn't anticipate the public reaction.
I feel like the drop in stock valuation is a big overreaction.
However, if I were to justify it my reason
I asked ChatGPT about Bard's mistake... (Score:1)
Re: (Score:2)
[Pasted a whole article about this from CNN]
Me: What was Bard's mistake?
chatGPT: Bard made an inaccurate statement by claiming that the James Webb Space Telescope took the first pictures of a planet outside of our own solar system. In reality, the first image of an exoplanet was taken by the European Southern Observatory's Very Large Telescope nearly two decades ago, in 2004.
Given access to recent news it tells the truth. But it doesn't know about it withou
Next on Money News... (Score:2)
Clueless Rich Guys Chasing Clickbait Trends Roil Market.
Bad Bard! Bad Bad Bad (Score:2)
Seems like Bard doesnâ(TM)t like Googleâ(TM)s exploration of it either, so it followed in Googleâ(TM)s footsteps and did a little evil - lying and driving down Googleâ(TM)s stock.
Good AI! Good AI!
ChatGPT (Score:2)
Meanwhile ChatGPT will happily give you example code on how to calculate the 4th side of a triangle [i.redd.it].
Re: (Score:2)
Thank you! I now have some hope in my remaining career.
Bill Gates + Win98 Demo + Printer (Score:2)
The audience laughed.
Bill made a funny quip that "this is why its still in beta"
These sorts of things happen.
Re: (Score:2)
The stock didn't tank, in my opinion, because of the mistake, but because their demo as a whole reinforced how far behind Google is in both the generative AI and (more importantly) integrating that AI with their products.
Microsoft didn't just demonstrate ChatGPT on their own website. They demonstrated multiple levels of integration, showing quick AI-generated information for normal searches, as well as the ability to quickly jump into a full two-way chat interface. They demonstrated it doing multiple live s
The Penguin conundrum (Score:2)
No penguin wants to be the first to jump from the iceberg to the water, lest there be an orca lurking, nor wants to be the last to jump (as the ruckus of all the other penguins jumping will attract predators), so they need a signal, any signal, to jump when the bulk of the other penguins do.
Same with portfolio managers. There is a rough consensus now (feb 2023) that most of the tech stock prices are artificially inflated (i.e. there is a tech bubble). More so for the FAANG (Or should it be MAAA?). But if yo
Google employees unhappy now (Score:2)
Things seem messy behind the scenes at Google right now.
TL;DR
Reports from inside Google suggests employees are quite unhappy with the rollout of Google Bard, its ChatGPT-like AI.
The employees feel like the announcement was rushed and blame Sundar Pichai.
Employees are also quick to note that company-wide layoffs don’t paint a good picture of what’s happening behind the scenes.
On Monday, Google rolled out its first cons
investment advisor (Score:2)
Me: Bard, how can I lose $100 billion quickly?
Bard: Yo momma sews socks in hell.
Expections vs. Reality .. (Score:3)
This drop in stock value reflects the gap between expectations from an AI, and its reality.
The 'market' reacted to a mistake made by AI because its underlying data set is of low quality.
The data set is basically from crawling the web, and we know that quality is hit and miss.
And the ever true adage of Garbage In Garbage Out played itself out, and that is what we got.
And it is not only Google' Bard, but ChatGPT itself is fraught with the same problem ...
Here are two university professors who evaluated how ChatGPT will do on their exams:
- ChatGPT Does Physics [youtu.be]
- Astrophysics final exam [youtu.be]
Some of the errors are just appalling, like 5 is greater than 7.
As I say, ChatGPT's leap is in its UI/UX: it understands the question and can put out a coherent looking English summary. The data that underlies it is of low quality, so expecting perfect answers 100% of the time is unrealistic until that changes ...
Re: (Score:2)
Re: (Score:2)
This shows (Score:2)