Hacker News Board Hacker News, browsed like a classic message board
Live data from Hacker News

Reflections on OpenAI

calv.info

Post reply on HN
Print view View original

241–250 of 446 posts

Re: Reflections on OpenAI

#241

Post by didibus »

ivape wrote: post #204
echoangle wrote: post #187

Earlier quoted context omitted.

> There was nothing hypothesized that next-token prediction and scale could show emergent properties. Nobody ever hypothesized it before it happened? Hard to believe.

Someone else can confirm, but from my understanding, no they did not know sentiment analysis, reasoning, few shot learning, chain of thought, etc would emerge at scale. Sentiment analysis was one of the first things they noticed a scaled up model could generalize. Remember, all they were trying to do was get better at next-token prediction, there was no concrete idea to achieve "instruction following", for example. W…

Didn't it just get better at next token prediction? I don't think anything emerged in the model itself, what was surprising is how really good next token prediction itself is at predicting all kind of other things no?

Re: Reflections on OpenAI

#242

Post by alwa »

ZYbCRq22HbJ2y7 wrote: post #234
alwa wrote: post #46

Earlier quoted context omitted.

If anyone tried to demand that I work that way, I’d say absolutely not. But when I sink my teeth into something interesting and important (to me) for a few weeks’ or months’ nonstop sprint, I’d say no to anyone trying to rein me in, too! Speaking only for myself, I can recognize those kinds of projects as they first start to make my mind twitch. I know ahead of time that I’ll have no gas left the tank by the end, and…

I think any reasonable manager would appreciate that sort of interest in a project and would support it, not demand it.

And, at least if I'm your manager, zealously defend the sanctity of your recovery time afterward.

Re: Reflections on OpenAI

#243

Post by nikcub »

mlinhares wrote: post #115
theletterf wrote: post #112

For a company that has grown so much in such a short time, I continue to be surprised by its lack of technical writers. Saying docs could be better is an euphemism, but I still can't find fellow tech writers working there. Compare this with Anthropic and its documentation. I don't know what's the rationale for not hiring tech writers other than nobody suggesting it yet, which is sad. Great dev tools require great doc…

The higher ups don't think there's value in that. Back at DigitalOcean they had an amazing tech writing team, with people with years of experience, doing some of the best tech docs in the industry, when the layoffs started the writing team was the first to be cut. People look at it as a cost a and nothing else.

I didn't realise that team at DO was let go, what a horrible decision - the SERP footprint of DO was immense and the quality of the content was fantastic.

Re: Reflections on OpenAI

#244

Post by yoyohello13 »

stickfigure wrote: post #231
rrrrrrrrrrrryan wrote: post #121

Earlier quoted context omitted.

> There's no Bond villain at the helm. It's good people rationalizing things. I worked for a few years at a company that made software for casinos, and this was absolutely not the case there. Casinos absolutely have fully shameless villains at the helm.

Interesting. A year ago I joined one of the larger online sportsbook/casinos. In terms of talent, employees are all over the map (both good and bad). But I have yet to meet a villain. Everyone here is doing the best they can.

I’m sure people working for cigarette companies are doing the best they can too. People can be good individuals and also work toward evil ends.

Re: Reflections on OpenAI

#245

Post by TheDong »

smikhanov wrote: post #182
kaashif wrote: post #159

Earlier quoted context omitted.

> I also don't believe AGI is a thing Why not? I don't think we're anywhere close, but there are no physical limitations I can see that prevent AGI. It's not impossible in the same way our current understanding indicates FTL travel or time travel is.

If we were to believe the embodiment theory of intelligence (it’s by far not the only one out there, but very influential and convincing), this means that building an AGI is an equivalent problem to building an artificial human. Not a puppet, not a mock, not “sorta human”, but real, fully embodied human, down to gut bacterial biome, because according to the embodiment theory, this affects intelligence too. In this fo…

Sure, if we redefine "AGI" to mean "literally cloning a human biologically", then AGI suddenly is a very different problem (mainly one of ethics, since creating human clones, educating, brainwashing, and forcing them to respond to chat messages ala chatGPT has a couple ethical issues along the way).

I don't see how claiming that intelligence is multi-faceted makes AGI (the A is 'artificial' remember) impossible.

Even if _human_ intelligence requires eating yogurt for your gut biome, that doesn't preclude an artificial copy that's good enough.

Like, a dog is very intelligent, a dog can fetch and shake hands because of years of breeding, training, and maybe from having a certain gut biome. Boston Dynamics did not have to understand a single cell of the dog's stomach lining in order to make dog-robots perfectly capable of fetching and shaking hands.

I get that you're saying "yes, we've fully mapped the neurons of a fruit fly and can accurately simulate and predict how a fruit fly's brain's neurons will activate, and can create statistical analysis of fruit-fly behavior that lets us accurately predict their action for much cheaper even without the brain scan, but human brains are unique in a way where it is impossible to make any sort of simulation or prediction or facsimile that is 'good enough' because you also need to first take some bacteria from one of peter thiel's blood boys and shove it in the computer, and if we don't then we can't even begin to make a facsimile of intelligence". I just don't buy it.

Re: Reflections on OpenAI

#248

Post by metaltyphoon »

chribcirio wrote: post #165
jordanmorgan10 wrote: post #155

I’m at a point my life and career where I’d never entertain working those hours. Missed basketball games, seeing kids come home from school, etc. I do think when I first started out, and had no kiddos, maybe some crazy sprints like that would’ve been exhilarating. No chance now though

> I’m at a point my life and career where I’d never entertain working those hours. That’s ok. Just don’t complain about the cost of daycare, private school tuition, or your parents senior home/medical bills.

> Just don’t complain about the cost of daycare, private school tuition, or your parents senior home/medical bills.

How does any of this relates to the amount of hours one works?

Re: Reflections on OpenAI

#249

Post by energy123 »

dcreater wrote: post #158

This is silicon valley culture on steroids: I really have to question if it is positive for any involved party. Codex almost has no mindshare and rightly so. It's a textbook also ran, except it came from the most dominant player and was outpaced by Claude code on the order of weeks. Why go through all that? Instead what would have been a much better scenario is openai carefully assessing different approaches to agent…

What I read in this blogpost is a description of how every good research organization works, from academia to private labs. The command and control, centrally planned approach doesn't work.

Re: Reflections on OpenAI

#250

Post by stickfigure »

yoyohello13 wrote: post #244
stickfigure wrote: post #231

Earlier quoted context omitted.

Interesting. A year ago I joined one of the larger online sportsbook/casinos. In terms of talent, employees are all over the map (both good and bad). But I have yet to meet a villain. Everyone here is doing the best they can.

I’m sure people working for cigarette companies are doing the best they can too. People can be good individuals and also work toward evil ends.

I am of the opinion that the greatest evils come from the most self-righteous.
Post reply on HN