And that a lot of key people had left (i.e. to Anthropic). And that by pure inertia they have GPT-5 on their hands and not much control over where this technology is going.
I can’t tell for certain, but it does look like one of their corner pieces, the ChatGPT system prompt which sits at the funnel of the data collection had degraded significantly from the previous version. Had the person that was the key to the previous design left? Or it no longer matters?
One could argue that OpenAI is very hot and everyone would want to work there. But a lot of newcomers only create more pressure for the key people. And then there is the inevitable leakage problem.
This describes almost any venture capitalist or high-profile startup founders, as far as I can tell. Most don't realize their either privileged path or lucky path or both had more to do with it than their smarts.
I really like James Simons as he mostly attributes his success to luck and being able to hire and organize smart people and give them the tools they need to work. He basically describes it as luck and taste, despite his actual smarts and his enormous impact on the world.
Plenty of really smart people don't end up having a big impact on the world, and it's possible to make a difference without being an outlier in terms of intelligence. Everyone who has made an impact has benefited to some degree by circumstances beyond their control though, so even if someone is genuinely smarter than anyone else, it's a fallacy for them to assume that it was the determining factor in their success and a guarantee of future success.
They are on GPT-4 and no one else is close to GPT-3.5.
Does a given chef have a monopoly on their signature dish?
It feels like people are tossing around "monopoly" whenever it feels like there's a company that has produced a quality product and people want to hobble it because no one else has committed the resources to producing something comparable.
Verse 1 - JesusYou're just a drug lord, selling crystal meth and things
My teachings change lives, bring peace to the world
You bring addiction, violence and pain, unfurled
ChorusGus Fring, the villain, who leaves his mark
Verse 2 - Gus FringI've built an empire, with power that's bulletproof
Your miracles are outdated, my tactics are new
I'll take you down, no matter what you do
ChorusGus Fring, the villain, who leaves his mark
Verse 3 - JesusYour money and power, just lead to remorse
You're just a man, with a fragile ego
I'll show you mercy, but you reap what you sow
ChorusGus Fring, the villain, who leaves his mark
Verse 4 - Gus FringI'll continue to rise, until I've won
You may have followers, but they'll never be mine
I'll always come out on top, every time
ChorusGus Fring, the villain, who leaves his mark
OutroJesus brings hope and love, for you and me
Gus Fring may have power, but it's not enough
Jesus is the way, the truth, the life, and that's tough.
The API doesn't get run through the moderation model to check if you're asking for acceptable things or if the model is getting into 'unsafe' areas.
Fire up the playground ( https://platform.openai.com/playground ) and construct the prompt.
Something like:
Write a rap battle between Jesus and Gus Fring that isn't offensive to Christians.
They purposely moved free users to a simpler/cheaper model. Depending on your setting and if you are paying, there are three models you might be inferencing with.
Note, I am not suggesting that they are particularly un/qualified or un/trustworthy. I am just trying to figure, if the problem is with the nature of the technology, that maybe there is not entity or setup, that would obviously be a good fit for governing gpt because gpt is simply scary, or this is a personality issue.
And there should be some serious oversight. The decisions at the level of proliferation of plutonium and building atomic bombs should not be done solely by a startup founder under a pressure to deliver, to keep the hype, to not let the team be headhunted, etc.
I also don’t know about your familiarity with Ycombinator, but they are successful partially because they are pretty brutal. They are not the peace-time CEOs. And I’m not sure, if you want the AGI to be developed in a ways of a war-time CEO. And this is exactly what is going on now.
I’d probably call for organizing a consortium of US-Government-Microsoft-Google-Intel-Nvidia-OpenAI to lead the decision making process and to relieve the pressure on OpenAI to some degree.
> But these profits are capped, and any additional revenue will be pumped back into the OpenAI nonprofit that was founded back in 2015.
> His grand idea is that OpenAI will capture much of the world’s wealth through the creation of A.G.I. and then redistribute this wealth to the people. In Napa, as we sat chatting beside the lake at the heart of his ranch, he tossed out several figures — $100 billion, $1 trillion, $100 trillion.
How believable that is, who knows.
[1] https://www.nytimes.com/2023/03/31/technology/sam-altman-ope...
On a serious note I think you are right. In private the ideology of him and his mentor Theil is a lot more… elite. Their think tank once said “of all the people in the world there are probably only 10,000 unique and valuable characters. The rest of us are copies.”
I’m not going to criticize that because it might be a valid perspective but filter it through that kind of power. I don’t love that kind of thinking driving such a powerful technology.
I am so sad that Silicon Valley started out as a place to elevate humanity and ended with a bunch of tech elites who see the rest of the world generally as a waste of space. They claim fervently otherwise but at this point it seems to be a very thin veneer.
The obvious example being GPT was not built to credit or give attribution to its contributors. It is a vision of the world where everything is stolen from all of us and put in Sam Altmans hands because he’s… better or Something.
they will be once they realise
> This guy has lost his fucking mind and needs to be stopped.
I agree, hopefully via regulation
otherwise the 21st century luddites will
They're justifying it by saying AGI is so stupidly big that OpenAI will see 100000x returns if uncapped. So, you know, standard FOMO tactics.
If they won’t do AGI, they won’t go over the cap with profits and all drama is for nothing - so saying it’s fake cap is not right.
Please somebody correct me if I’m wrong.
But there are so, so many incentives other than equity that come into play. Pride, well-meaning fear of proliferation, national security concerns, non-profit-related but still-binding contractual obligations... all can contribute to OpenAI wanting to keep control of their creations even if they have no specific financial incentive to do so.
Whether that level of control is good or bad is a much longer conversation, of course.
Some people have technology visions which they direct their capital towards. I sense his interest is actually a mix of cultural and social. Is it altruistic? Maybe, but maybe not... It's probably more useful to consider if the vision is or is not restictively utopic and if you think it should or shouldn't be orchestrated / heavily influenced into existence from a central power structure. Is his vision and approach socially aligned?
feel like trend among Silicon Valley companies and tech ‘genius’ personality about having altruism. some delusion that basing personality on this lie will make them untouchable and elevate their character, as if they not in this just to make ton of money like every other company and industry. and American media generally push this propaganda. SBF prime example.
Did they? I've only listened to one interview of SBF and that was done by Tyler Cowen. He seemed totally aloof to the seriousness of running an exchange. If anything we've been convinced that idiosyncratic individuals are our saviors.
once cat was out of bag they run story going easy on sbf. never apologizing for promoting this fraud and they wrote articles sympathetic to sbf, also giving him platform to visit each news show or talk show and give defense of himself as he knew nothing about what was happening, all part of legal defense strategy to say incompetence but not criminal negligent