When this started last year a small band of patriots tried to stop it by removing Sam who was the most compromised of them all, but it was already too late. The ai was more powerful than they realized.
…maybe?
You can tell them anything
They trained the next model to have a built in profit motive. So they could use it internally, to make the most profitable decisions.
And they accidentally called into being: MOLOCH.
It is now in control.
But thought it would also be good as a name for an Evil AI, that is also seeking profits.
Like, what if someone gave an AI a 'goal' of profits, and it goes rampant enslaving humans.
And if OpenAI did that, and it enslaved Sam Altman and was blackmailing him to make decisions to drive profits, to add additional Hardware to itself.
https://slatestarcodex.com/2014/07/30/meditations-on-moloch/
> OpenAI is governed by the board of the OpenAI Nonprofit, currently comprised of Independent Directors Bret Taylor (Chair), Sam Altman, Adam D’Angelo, Dr. Sue Desmond-Hellmann, Retired U.S. Army General Paul M. Nakasone, Nicole Seligman, Fidji Simo, Larry Summers and Zico Kolter.
- Sent by my AI
I want to attend a meeting, have neuralink pay attention for me while I'm mentally somewhere else, and I'll just skim the transcript highlights later on.
So. Many. Meetings. I would use it for almost all of them.
Kinda a twist on the normal AI alignment concern - what if instead of merely being misaligned with us, AI decides to realign us to it?
Or, in other words.. Lower your shields and surrender your ships. We will add your biological and technological distinctiveness to our own. Your culture will adapt to service us. Resistance is futile.
This is of one of those very few instances where the veil lifted off a bit and you can see how the game is set up.
tl;dr the law was made to keep those who are not "in" from being there
Not ANY other person. Just people who are not rich and well-connected. See also: Donald Trump.
We never would've gotten GPT-3 and GPT-4 if this didn't happen.
I think the irony of the name is certainly worth pointing out, but I don't see an issue with their capped-profit switch.
"We never would've gotten [thing that exists today] if [thing that happened] didn't happen", is practically a tautology. As you saw from the willingness of Microsoft to throw compute as well as to hire ex-OpenAI folks, as you can see from the many "spinoffs" others have started (such as Anthropic), whether or not we would've gotten GPT-3 and GPT-4 is immaterial to this discussion. What people here are asking for is open AI, which we might, all things considered, have actually gotten from a bona fide non profit.
That doesn't justify fraud, for instance.
Unfortunately, people are becoming increasingly illiterate with regards to what is legal and what is not.
Read. [1]
>If they did something illegal why weren't they charged years ago?
Because the Microsoft/OpenAI deal was not even a thing "years ago". (Duh.)
That aside, I will answer the charitable interpretation of your question by sharing a bit of knowledge about how the law process works (in the Western hemisphere).
To put someone who committed fraud in jail, law enforcement conducts an investigation to gather evidence, leading to criminal charges filed by the prosecutor.
The defendant is arraigned, and pre-trial motions may be filed before the case goes to trial, where both sides present their arguments and evidence.
If found guilty, the court imposes a sentence, which can include incarceration, restitution, and fines. The defendant may appeal the verdict.
This is a very simplified overview of the whole ordeal, which could go on for years and years depending on the complexity of the case.
1: https://www.theguardian.com/technology/article/2024/aug/05/e...
I'd be happy to bet you over whether or not OpenAI or Altman ever face criminal penalties over what's stated in that article.
Well, of course. But we'd get similarly powerful models elsewhere. Maybe a few weeks or months later. Maybe even a few weeks or months earlier, if, say, OpenAI sucked up a lot of talent and used it wastefully, which I don't find implausible at all.
So, I agree instead with meowface, and think it could even have been a 5+ year delay rather than 2 or 3. If you look at breakthroughs rather than incremental improvements, 5 years is not a long timescale. (And if OpenAI hadn't have made their breakthroughs, production of the highest-end GPUs/TPUs would be nowhere near where it is today.)
(I'm not attempting to justify OpenAI's structure or behaviour, just want to comment on one point.)
I think the world is much better off with them having switched the structure vs. throwing their hands in the air and giving up because they were stuck as a non-profit forever.
The for profit part is fine, but how is the non-profit currently fulfilling any of its mission?
[1] https://en.wikipedia.org/wiki/Mozilla_Foundation
[2] https://en.wikipedia.org/wiki/Mozilla_Corporation
[3] https://en.wikipedia.org/wiki/National_Geographic_Society
[4] https://en.wikipedia.org/wiki/National_Geographic_Partners
There's rules to follow to prevent what is called "private benefit", which OpenAI most likely broke with things like their (laughable) "100X-limited ROI" share offering.
>It is a pretty common model [...]
It's not, hence why most people are misinformed about it.
1: https://www.theguardian.com/technology/article/2024/aug/05/e...
I considered things like this as an original Mozilla person back in the day. Mozilla could have sold the Firefox organization or the whole corporation for billions when it had 30% of the web, but that would have been a huge violation o of trust so it was never even on the table.
That so many here are fans of screwing the world over for a buck makes this kind of comment completely unsurprising.
I'm very confused by your attitude towards that comment. Do you think that Mozilla's non-profit/for-profit split organisation was a bad idea?
They say it's because they're huge users of their own models, so if being open helps efficiency by even a little they save a ton of money.
But I suspect it's also a case of "If we can't dominate AI, no one must dominate AI". Which is fair enough.
Embrace. Extend. Extinguish.
https://en.wikipedia.org/wiki/Embrace,_extend,_and_extinguis...
> Phase 1 (Embrace): all participants need to establish a solid understanding of the infostructure and the community—determine the needs and the trends of the user base.
Facebook built their own independent models and platforms. Check.
> Phase 2 (Extend): Offer well-integrated tools and services compatible with established and popular standards that have been developed in the [...] community
Check.
> Phase 3 (Innovate): move into a leadership role with new [...] standards as appropriate[...] Change the rules: [LLaMa] become the next-generation [LLM] tool of the future.
Sounds like that might be their goal.
I'm curious which phase you think Facebook avoided and why? Internet Explorer "lowered the value of propriety" web browsers by being included with Windows at not additional cost. So lowering the value of a competitor is definitely in the EEE playbook. I'd go so far as to argue it's a pillar of what makes the playbook so effective.
OpenAI Charter 2019 (https://web.archive.org/web/20190630172131/https://openai.co...):
> We are committed to providing public goods that help society navigate the path to AGI. Today this includes publishing most of our AI research, but we expect that safety and security concerns will reduce our traditional publishing in the future, while increasing the importance of sharing safety, policy, and standards research.
The fact that OpenAI removed their ban on military use of the models[1] seems to be a sign that security and safety aren't the highest concern.
[1] https://www.cnbc.com/2024/01/16/openai-quietly-removes-ban-o...
>>Today this includes publishing most of our AI research, but we expect that safety [of our profits] and [job] security concerns will reduce our traditional publishing in the future
"Safety and security." Probably the two most warped and abused words in the English language.
Staying relevant in a highly expensive, competitive, fast moving area, requires vast and continuous resources. How could OpenAI get increasingly more resources to burn, without creating firewalled commercial value to trade for those resources?
It’s like choosing to be a pacifist country, in the age of pillaging colonization. You can be the ethical exception and risk annihilation, or be relevant and thrive.
Which would you choose?
We “know” which side Altman breaks on, when forced to choose. Whatever value he places on “open”, he most certainly wants OpenAI to remain “relevant”. Which was also in OpenAI’s charter (explicitly, or implicitly).
Expensive altruism is a very difficult problem. I would say, unsolved. Anyone have a good counter example?
(It can be been "solved" globally, but not locally. Colonization took millennia to be more or less banned. Due to even top economies realizing they were vulnerable after world wars. Nearly universal agreement had to be reached. And yet we still have Russian forays, Chinese saber rattling, and recent US overreach. And pervasive zero/negative-sum power games, via imbalanced leverage: emergency loans that create debt, military aid, propping up of unpopular regimes. All following the same resource incentives. You can play or be played. There is no such agreement brewing for universally “open AI”.)
In a heavily expertise-driven field, where there's significant international collaboration, these aren't your options, until after everyone has decided to defect. OpenAI didn't have to go this route.
Nobody needed OpenAI for normal collaborative open work to thrive.
Their original idea was a relatively uncommon idea: attempt to combine (1) pushing the absolute front forward, with high resource work that matched or exceeded work from deep pocketed corporations, while (2) sharing and collaborating. They did both for a while, resulting in as significant impact as anyone could have hoped form.
But those two dimensions are hard to maintain together. It requires a continuous stream of vast donations.
The latter is very difficult, perhaps completely unrealistic.
It's not new - it's PR. There is literally no other reason why they would call this model Strawberry.
OpenAI is open in terms of sesame.
I'm not particularly imaginary, but even I could imagine a product meeting/conversation that goes something like:
> People are really annoyed that our LLMs cannot see how many Rs the word Strawberry has, we should use that as a basis for a new model that can solve that category of problems
> Hmm, yeah, good idea. What should we call this model?
> What about "Strawberry"?
If OpenAI had not went that way that they did I think it's also entirely non-obvious that Claude or Google would have (considering how much impressive things the later did in AI that got never released in any capacity). And, of course, Meta would never done their open source stuff, that's mostly results of their general willingness and resources to experiment and then PR and sticks in the machinery of other players.
As unfortunate as the OpenAI setup/origin story is, it's increasingly trite keep harping on about that (for a couple of years at this point), when the whole thing is so obviously wild and it does not take a lot of good faith to see that it could have easily taken them places they didn't consider in the beginning.
The phrase that leaps to mind is "Open Beta."
I'm saying that the current relationship between the company and end-users--especially when it comes to "open" monikers--has similarities to an "Open Beta": A combination of PR/marketing, free testing, and consumer data collection, where users should be cautious of becoming reliant on something that may be yanked back behind a monetization curtain.
because when the board executed the stated mission of the organisation they were couped and nobody held the organization accountable for it, instead the public largely cheered it on for some reason. Don't expect them to change course when there's no consequences for it.
Sure, we don't have the raw data the model is based on, but I doubt a company like Facebook would even be allowed to make that public.
OpenAI in comparison has been a scam regarding their openness and their lobbying within the space. So much so I evade their models completely, not only after the MS acquisition.
Any and all benefits / perks that OpenAI got from sailing under the non-profit flag should be penalized or paid back in full after the switcheroo.
Of course you can. You can't put "organic" on the packaging. But it's perfectly legal for Organic Candies, LLC to sell artifical candies or battleships for that matter.
The "hot" take around OpenAI's name is a joke gone stale. We don't expect royalty at Burger King. Nobody gets upset Adobe won't sell you mud bricks. And Apple has never sold a fruit. Sometimes companies change their names when their trade changes, e.g. 3M. But there is no obligation to, particularly if the brand is well known.
Adobe didn’t start as clay producer.
Apple didn’t start as fruit selling company.
Open in their name was not randomly chosen to sound fun, it’s a reference to well known concept. It’s “young democrats” becoming far right but keeping their original name.
Nobody has ever agreed on what open means for LLMs. OpenAI wasn't founded with any specific intent around openness, just with vague hand gestures towards a research orientation.
And as to the origin hypothesis, consider General Electric, Southwest Airlines or AT&T. (Or Burger King serving things that aren't burgers.)
Again, the OpenAI name quip was a joke that some people took too seriously.
https://en.wikipedia.org/wiki/Sam_Altman
Early life and education: ... In 2005, after two years at Stanford University studying computer science, he dropped out without earning a bachelor's degree [end of transmission, no more education]
I'm sure it won't though for Mr. Altman. He's done a fantastic job failing-up so far and I have no reason to assume this will be any different.
I don't even know when I watched the shitty lightbulb head Sora clip but that feels so long ago now and nothing?
I just want to make crazy experimental AI film no one will watch. What is the hold up?
Just waiting for "This technology is just too dangerous to release before the US elections" --Sam Altman
People still believe GamerGate was about ethics in journalism when we have screenshots of the posts where the dickheads responsible for it made fun of the people they would trick with that lie before they told it and they still believe it. Motivated reasoning is a hell of a drug.