OpenAI CEO Sam Altman on GPT-4 Hype: 'People are Begging to be Disappointed and They Will Be' (theverge.com) 46
The Verge writes:
OpenAI CEO Sam Altman has addressed rumors regarding GPT-4 — the company's as yet unreleased language model and latest in the GPT-series that forms the foundation of AI chatbot ChatGPT — saying that "people are begging to be disappointed and they will be." During an interview with StrictlyVC, Altman was asked if GPT-4 will come out in the first quarter or half of the year, as many expect. He responded by offering no certain timeframe. "It'll come out at some point, when we are confident we can do it safely and responsibly," he said....
When asked about one viral (and factually incorrect) chart that purportedly compares the number of parameters in GPT-3 (175 billion) to GPT-4 (100 trillion), Altman called it "complete bullshit."
"The GPT-4 rumor mill is a ridiculous thing. I don't know where it all comes from," said the OpenAI CEO. "People are begging to be disappointed and they will be. The hype is just like... We don't have an actual AGI and that's sort of what's expected of us."
Asked about how far we are from developing AGI, Altman replied "The closer we get, the harder time I have answering. Because I think it's going to be much blurrier and much more of a gradual transition than people think."
And Altman also addressed predictions that ChatGPT will kill Google. "I think whenever someone talks about a technology being the end of some other giant company, it's usually wrong. I think people forget they get to make a countermove here, and they're like pretty smart, pretty competent. I do think there's a change for search that will probably come at some point — but not as dramatically as people think in the short term."
When asked about one viral (and factually incorrect) chart that purportedly compares the number of parameters in GPT-3 (175 billion) to GPT-4 (100 trillion), Altman called it "complete bullshit."
"The GPT-4 rumor mill is a ridiculous thing. I don't know where it all comes from," said the OpenAI CEO. "People are begging to be disappointed and they will be. The hype is just like... We don't have an actual AGI and that's sort of what's expected of us."
Asked about how far we are from developing AGI, Altman replied "The closer we get, the harder time I have answering. Because I think it's going to be much blurrier and much more of a gradual transition than people think."
And Altman also addressed predictions that ChatGPT will kill Google. "I think whenever someone talks about a technology being the end of some other giant company, it's usually wrong. I think people forget they get to make a countermove here, and they're like pretty smart, pretty competent. I do think there's a change for search that will probably come at some point — but not as dramatically as people think in the short term."
What's This? (Score:2, Insightful)
A reasonable, measured CEO? That won't go over well with the 400x return people.
Re:What's This? (Score:5, Insightful)
A reasonable, measured CEO?
OpenAI Inc is a non-profit.
Since he doesn't answer to shareholders, he can speak the unfiltered truth.
Most CEOs can't do that. They would be crucified. Their job is to twist, distort, and hype the truth.
Re: (Score:2)
A reasonable, measured CEO?
OpenAI Inc is a non-profit.
Since he doesn't answer to shareholders, he can speak the unfiltered truth.
Most CEOs can't do that. They would be crucified. Their job is to twist, distort, and hype the truth.
We need way more non-profit corporations in the tech world. And in everything else, really, but particularly in tech.
Re: (Score:2)
We need way more non-profit corporations in the tech world.
Perhaps. But OpenAI doesn't produce any products. It conducts research. So it serves a different purpose than for-profit tech companies.
It is also funded very differently. For-profit corporations are funded by investors and customers. OpenAI is funded by donations from billionaires who made their money from for-profit corporations.
Re: (Score:2)
developer.
chat g p t is cool to use.
client.
oh.
alright.
i want it to look blue
Re: (Score:2)
We need way more non-profit corporations in the tech world. And in everything else, really, but particularly in tech.
Unfortunately for your point of view, Open AI had to become for-profit because it couldn't obtain enough funding otherwise. There are simply too many ways for capital to make large returns for a non-profit to obtain the funding necessary to do big things (like cutting edge AI research).
That isn't true for everything, but it appears to be true for most things.
Re: What's This? (Score:1)
Re:What's This? (Score:5, Informative)
Weeellll:
"OpenAI is an American artificial intelligence research laboratory consisting of the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc."
Re: (Score:2)
OpenAI is an American artificial intelligence research laboratory consisting of the for-profit corporation OpenAI LP and its parent company, the non-profit OpenAI Inc."
Yes... but it is a nonprofit which owns all the shares, so the shareholders aren't chasing them for profits in the same way. I gather that the reason for this structure is that non profits are a pain in the arse to do accounting for plus have all sorts of rules. But they can own profit making companies and have one very large line item of "fu
Re: (Score:2)
Re: (Score:2)
100T sized model is a stupid idea (Score:2)
Who would pay so much!!?? See how stupid would be to have a 100T model today?
Re: (Score:3)
A human brain has about 1500T, although it runs much, much slower (about 50 Hz).
Re: 100T sized model is a stupid idea (Score:2)
It's also structurally and operationally far more complex.
Non BS CEO (Score:5, Informative)
Re: (Score:3)
Sorry, but GPT *is* an AI, it's just not an AGI. It's only got one domain of expertise, and, more to the point, it can't learn new domains of expertise.
OTOH, I would assert that humans aren't real AGIs. There are thoughts too complex for them to handle. (There's variation, but not a lot.) An artificial AI could, in principle, get a lot closer to being a real AGI than humans can.
OTOH, each increment of intelligence seems to have progressively narrowing areas of application. I've a theory, with no eviden
Re: (Score:2)
Unless you believe we are living in a simulation
Congratulations. Someone believes this may be the case and they're working to prove/disprove this idea [popularmechanics.com].
Re: (Score:2)
Well, you've got a point, but not the one I intended. But if I'd denied that humans were GI's, you'd argue about that, too. And my point was that human intelligence isn't really general, even though it's a lot broader than any other examples we know of.
Re: (Score:2)
Re: (Score:2)
I don't actually believe this, but sooner or later we're going to get something that understands how to lie.
Re: (Score:2)
There's more to it. It 'understands' questions well enough to answer them. Likewise, critiques. Likewise, coding.
And what it doesn't understand, it BS's about incredibly well.
Re: Sam Altman: "We don't have an actual AGI". (Score:2)
I agree. I found it was pretty good at coding, and this corresponds to 40,000 tech layoffs THIS WEEK:
https://www.wsws.org/en/articl... [wsws.org]
Pinching Pennies? Odd MS dropped a million on a private sting concert the night before:
https://fortune.com/2023/01/20... [fortune.com]
The reality is they are no longer needed. I think big shake-ups coming.
Re: (Score:2)
It returns a match from its database.
It is a Chinese Room.
It is not intelligent.
Re: (Score:2)
Re: (Score:2)
Yeah, that isn't at all what it does. Unfortunately, the people who understand it keep explaining it, but you keep not listening.
Re: (Score:2)
the people who understand it keep explaining it
Any good sources?
I asked ChatGPT how it works:
Me: How do you understand what I'm writing?
ChatGPT: I understand what you're writing by processing the text you provide as input and using my pre-trained language understanding capabilities to generate an appropriate response. This process is based on machine learning techniques that allow me to understand the context and intent of the text and generate a response that is relevant to that context and intent.
Me: How do you represent context and intent internally?
Re: (Score:2)
Here you go, source code so you can train your own; a manageable size, but same structure: https://github.com/karpathy/mi... [github.com]
You will note the absence of a word database in the model.
Re: (Score:2)
I came across this today and remembered your post. You might find it interesting. I just skimmed it, but it seems okay, and the diagrams are decent. It's GPT-2 and predecessors, but GPT-3 is very similar but bigger. My original reply has the real primary sources though.
https://jalammar.github.io/ill... [github.io]
Asked about how far we are from developing AGI (Score:4, Funny)
He laughed and laughed.
We later asked a parachute manufacturer how far we are from developing anti-grav boots. She gave the same answer.
Re: (Score:2)
She gave the same answer.
And her great ^ (10E99999) grandchildren will still be laughing along with their language model peers. AGI is a pipe dream for non-biological entities. Some would argue that it's also a pipe for human entities, and there's a strong argument to be made for that opinion.
Anti-gravity boots are also a pipe dream, by the way. Some form of rocket boots seems plausible, but I'd put them at 50+ years from now to be at all practical.
TIL: Elon Musk co-founded OpenAI (Score:2)
(*) qv His Pelosi tweet
Re: TIL: Elon Musk co-founded OpenAI (Score:2)
Err, no. Electric cars and spaceships really arnt affecting the lives of the vast majority of the population at all.
An example of a non politician who has seriously changed the world then you could do worse than suggest Tim Berners Lee.
Re: (Score:2)
On Musk's side:
- Global warming could end humanity.
Transportation accounts for 27% of total U.S. greenhouse gas emissions.
- Spacex's Starship will transform access to space.
It will make it possible for us to live on another planet and it will enable telescopes that could discover life on another planet.
- AI research might one day actually lead to actual Artific
Re: TIL: Elon Musk co-founded OpenAI (Score:2)
You have to drive an electric car a long way to make up for the cost of its manufacture in CO2 terms.
Access to space to go where? Living on other planets is a pipedream for people who dont understand thermodynamics or biology. If it was easy NASA would
have done it decades ago. The moon is barren and mars is a toxic desert. One breath of the perchlorates in its dust and your lungs would be finished.
The rest is all maybe, might, possibly.
CO2 and two cars. (Score:2)
You have to drive an electric car a long way to make up for the cost of its manufacture in CO2 terms.
Around here that's 2 years. Might be a bit longer in the US. It all depends on exactly how clean the considered country's electric mix is.
Still nearly most of the time, it's way earlier than the 8 years of the batteries' life time.
People tend to forget how many hectolitre of CO2-producing fuel a regular car burns_
- It's not EV are such low carbon to produce,
- is that ICE are such insane CO2 emitters that EV can quickly achieve parity despite their manufacturing.
Still if you want to actually have an impact o
Re: (Score:2)
"Around here that's 2 years."
2 years is not a distance. I've read in more than a few places than an EV needs to do more than 80K miles to start saving on CO2 compared to an ICE car.
"(Mars is the one with perchlorate, not the moon, just saying.)"
Hence I said Mars is a toxic desert.
Re: (Score:2)
I have heard it is quite cold there.
> The rest is all maybe, might, possibly.
No. It is not.
Re: (Score:2)
- Global warming could end humanity.
Transportation accounts for 27% of total U.S. greenhouse gas emissions.
...and to have a sufficient impact on the climate you can't merely switch car to electrical drive.
You'd need to dramatically switch transportation toward public transport and e-bike.
Which is definitely not even been considered in the US (unlike here around in Europe, where this could be a realistic long term goal).
- Spacex's Starship will transform access to space.
It will make it possible for us to live on another planet
Which is still a couple centuries (and several major shifts in technologies) away.
At which point, the fact that Space X was a player in the first couple of decade wouldn't matter that much. At bes
Re: (Score:2)
"He's just some rich twat with lots of money"
And you, presumably, are a rude twat who could have Musk's money, but chooses not to.
Got it.
Birth money (Score:2)
And you, presumably, are a rude twat who could have Musk's money, but chooses not to.
To have Musk's money would mostly require to be born from Musk's parent and thus inherit the necessary money to gamble on such tech investments.
Not everyone has won what Cory Doctorow calls the "lucky orifice lottery".
Though the latter don't prevent people having lots of fun in research in academia, while working on very big impact projects, like playing a critical role in jump-starting the variant tracking effort in my country and helping bring into production an entirely new approach to variant tracking t
Re: (Score:1)
Then you might be able to spout better drivel.
$1 billion (Score:2)
Is MS draining $1 Billion