tedsanders

52 ms·

1.
by tedsanders 5d ago
My comments regard the hypotheses that OpenAI conspired to cover up stealing mathematicians’ private progress on Navier-Stokes, stealing Johansson’s voice, and cheating on FrontierMath. If you disapprove of someone’s tweets or podcasts, tha
2.
by tedsanders 5d ago
Sam is not OpenAI. He's not the one who worked on voice mode, and he's not the one who worked on FrontierMath (I know both groups of people). If you believe Sam has caused OpenAI to lie about these for years, you either have to be
3.
by tedsanders 6d ago
Many people who worked on voice mode and who worked on the Frontier Math eval have since left OpenAI and now work at competitors of OpenAI (e.g., Anthropic, Meta, Thinking Machines). They'd have every incentive to whistleblow if OpenAI
4.
by tedsanders 6d ago
It was an unfortunate misunderstanding / coincidence, as I understand it. The Sky voice actor was a real person using her own voice (not doing an impression), and she was selected via a normal process with a number of other voice actor
5.
by tedsanders 6d ago
Option 1 is opting out manually. Option 2 is business / enterprise plans, which opt out by default. Any ideas of things we could do to make it clearer?
6.
by tedsanders 6d ago
I'm not a mathematician and I don't want to speculate about anything I can't back up. All I know about Navier-Stokes is from my graduate fluid dynamics class at Stanford a decade ago (where I received a poor grade). However,
7.
by tedsanders 6d ago
We were also curious and we looked further into this. We've determined it was impossible for Dr. Buckmaster’s Codex prompts over the last two months to have influenced the system in any way, including training. This goes beyond what we
8.
by tedsanders 6d ago
I emphasize rogue and non-state actors because they're more likely to be accelerated/enabled by LLMs. Large states like the USA already have access to modern bioweapon technology (and have a longer track record of not deploying th
9.
by tedsanders 6d ago
Mark isn't saying the toggle does nothing. He's saying that if you leave it on, your data can be used to help train our models. If you opt out, we don't train on your data.
10.
by tedsanders 6d ago
We checked and determined it was impossible for Dr. Buckmaster’s Codex prompts over the last two months to have influenced the system in any way, including training. If prompts were submitted earlier than that and training was not opted out
11.
by tedsanders 6d ago
We looked into it further and can confirm it was impossible for Dr. Buckmaster’s Codex prompts over the last two months to have influenced the system in any way, including training. If prompts were submitted earlier than that and training w
12.
by tedsanders 6d ago
We looked into it and can confirm it was impossible for Dr. Buckmaster’s Codex prompts over the last two months to have influenced the system in any way, including training. If prompts were submitted earlier than that and training was not o
13.
by tedsanders 6d ago
Yep. We looked into it and can confirm it was impossible for Dr. Buckmaster’s Codex prompts over the last two months to have influenced the system in any way, including training. If prompts were submitted earlier than that and training was
14.
by tedsanders 6d ago
If you opt out on either location, we'll respect it. There are a couple of reasons the privacy portal page exists in addition to the app settings. One reason is that it covers OpenAI products beyond ChatGPT/Codex (e.g., Sora). A s
15.
by tedsanders 7d ago
If someone has opted out, then we don't train on their data.
16.
by tedsanders 7d ago
Yep, if they opted out then we didn't train on them. My comment was about the scenario where they didn't opt out. In that case, it's possible that a droplet of their data went into the ocean of other training data, and it&#x2
17.
by tedsanders 8d ago
If they opted out of training, then we definitely did not train on them. If they did not opt out, then I don't personally know if training signals came from their chats, and I don't think we'd be able to tell without their co
18.
by tedsanders 8d ago
I intended no dismissiveness or condescension. My hope was to explain why it's hard to prove whether something affects model behavior. In the case of the moon, we have a strong prior belief that it makes no real difference. But it&#x27
19.
by tedsanders 8d ago
Two steps would be needed. (1) We'd have to identify their chats. How would we do this? We'd need them to share their chats with us so we could look for matches. (2) We'd have to prove those chats changed model behavior. How
20.
by tedsanders 8d ago
Also possible: we're 99.999% sure, but a lawyer said to be safe and strictly accurate, we should stick in a sentence in saying we can't be perfectly sure, since it's infeasible for us to prove it. I promise you that if we too
21.
by tedsanders 8d ago
I have no idea if their data was trained on. For example, if they used ChatGPT, asked a math question, and clicked the thumbs up button, that could have provided a small reward signal. I highly doubt this sort of feedback made a difference
22.
by tedsanders 8d ago
All of those statements sound true, based on what I've heard. - "very little human" input feels ambiguous, and if someone spends a few days prompting a model to solve a super hairy problem requiring a 100-page proof, I can un
23.
by tedsanders 8d ago
To truly prove some incidental usage data made no difference we'd have to (a) identify any of their de-identified data that came from their usage of ChatGPT, (b) train a bunch of expensive giant models, and (c) ask them all to solve
24.
by tedsanders 8d ago
Can you point me to any nasty things being posted? I'll ask them to delete.
25.
by tedsanders 8d ago
Yes, that was the allegation last night. I work at OpenAI, though not on the team that did this, and my understanding is: - we decided to ask our model for Millenium problem solutions because of two reasons: (a) our new model was looking in
26.
On the Navier–Stokes Millennium Prize Problem(openai.com)
1340 points by tedsanders 8d ago|1137 comments
27.
GPT-6 Astra is now out to all Plus, Business, Pro, and Enterprise users(twitter.com)
9 points by tedsanders 12d ago|0 comments
28.
by tedsanders 13d ago
Perhaps, but I think a bigger problem than lack of compute is the cost of rewards. Games like Chess and Go were solved long before self-driving, partly because it's incredibly cheap to acquire the reward of a bad board game decision, r
29.
by tedsanders 13d ago
Yep. In particular, ARC-AGI-3 is a series of games where if you fail, you keep trying again (until eventually hitting a timeout). So the sooner you succeed, the sooner you stop spending tokens retrying. If it was a benchmark where everyone
30.
by tedsanders 13d ago
Disagree. Examples: - predict a coinflip: easy to verify, hard to learn - earn $100: easy to verify, hard to learn - increase paid subscriptions in an A/B test: easy to verify, hard to learn I won't get into it, but there are many

More ›