Earlier quoted context omitted.
I've got access to GPT-4 at work. There have been many times that I'll encounter a bug while programming, I'll paste my code to GPT, tell it what my code is currently doing, and tell it what I actually want the code to do. It then explains what I'm doing wrong and what I can do to fix it. I've had success with this over 90% of the time. This saves me a significant amount of time that would have otherwise been spent h…
How is this possible? What do you guys do at work? I haven't had success with neither GPT-4 (did you build your own API calling tool for it? Do you just paste it in their Playground?) nor with GitHub Copilot in actually delivering anywhere close to 90% of the time. It usually misses a whole lot of context. It feels like it would work for perfectly encapsulated small single purpose functions, which of course sounds gr…
What it feels like to work in AI right now
671–680 of 768 posts
Re: What it feels like to work in AI right now
#672Earlier quoted context omitted.
I think this is a classic case of us overestimating the immediate impact and underestimating the long term impact. Right now, they are definitely useful time savers, but they need a lot of handholding. Eventually, someone will figure out how to get hundreds of LLMs supervising teams of millions of LLMs to do some really wild stuff that is currently completely impossible. You could spin up a giant staff the way we do…
>Eventually, someone will figure out how to get hundreds of LLMs supervising teams of millions of LLMs to do some really wild stuff that is currently completely impossible. This is an intuitive direction. In fact, it’s so intuitive that it’s a little bit odd that nobody seems to have made proper progress with LLM swarm computation.
Re: What it feels like to work in AI right now
#673Earlier quoted context omitted.
I've calmed down after your "heads will roll" comment so I promise to be civil. To continue. Yes, you should put back your original comment to reduce confusion. I'm fine reading a long comment. > No idea why you had to tout out that you had a PhD when I was pointing out that you were not a user. You said: And then we have bystanders like you with no understanding of what’s actually going on, who want to ban everybody…
Again I did in fact open pornhub when you point it out and that’s Tesla Autopilot not FSD. I do find it beyond the pale that people drunk use FSD, because there are very strict controls for people using FSD. Autopilot is advanced cruise control, apart from regulatory authorities saying it’s safe compared to competitors, Tesla releases quarterly safety reports since 2019 counting the incidents with Autopilot. It is sh…
So what did you do to fix the issue? Why did a second person die in the same exact way as the first person after 3 years? Were you working on the fix at all? Or did you do nothing? Just admit it if you did nothing in response to that decapitation.
You say "It's Autopilot, not FSD" as if that absolves Tesla of anything. Your marketing does not change my opinion of your technologies. It doesn't change that FSD and AP both have glaring technological flaws, doesn't change the fact that FSD is beta-quality hardware and software being tested on the general public, something the public did not agree to. It doesn't change the fact that even though they didn't agree to it, you unilaterally decided it was okay to conduct a beta test involving us. That's a huge ethical problem, and the fact you don't even see it as such blows a hole in your insistence that your colleagues take safety seriously.
It doesn't matter. I understand why you think it absolves you; because you feel that the technology is similar enough to others out there, that it's just an incremental step, and so how can Tesla be held responsible when people make a career out of fucking in their car using that technology? How can Tesla be held responsible when multiple people lose their heads due to poor choices in sensor design? Correct me if I'm off base. Why is the AP/FSD distinction so important to you?
How do you not realize it's your entire company's fault there was no other orthogonal sensor to see the tractor trailer? How do you not realize it's your entire company's fault people out there feel safe enough to use your product to watch Harry Potter or fuck while while flying down the highway, using what you call "glorified cruise control"? That's what you say it is to me, your peer, but to them you've said it was "AutoPilot (TM)". Why didn't you call it "Glorified Cruise Control" or just "Cruise Control"?
That's not on them, that's on you for unleashing this technology on us. People are always going to watch movies and fuck. That they're doing so in your beta-quality robot menace to society is not their fault. Your company specifically conditioned them to think it was okay to do this in a Tesla.
> It is shown to be safer than humans behind the wheel by quite a margin.
The passive voice is doing a lot of work here.
> There is a risk of users being too careless with FSD, which is why people we put quite some effort to getting rid of such users quickly.
Again, this just goes to show how Tesla, in fact, is not concerned about safety and security, but instead are laser focused on market dominance and pushing technology on us as fast as possible. Tesla is a look before you leap, shoot first and ask questions later kind of company. Or as I said, move fast and break things (or in this case, "put quite some effort" to patch them up after the fact).
This is a brand new technology and Tesla is rushing it out to the public as fast as humanly possible, selling it in beta quality before the technology and software is even ready. And you're telling me now that you screen users for bad behavior and ban them after the fact. It just goes to show you're treating this as some grand social experiment you feel you have the right to run on the rest of us.
Yes people are going to watch movies in their cars. They're going to fuck. Yes they will be drunk and asleep. The problem is that you don't seem to care that your customers are using your products to do all these things, and it's over the course of years.
Re: What it feels like to work in AI right now
#674AI world can be proper dystopian these days. I know someone who accepted a job offer in one of the biggest market research firms in the world. She was making AI models for a high tech company before and was hired to make summarization AIs. Between serving out the (frankly ridiculous) 2 month notice period and joining this new job GPT-4 was launched and prices came down for 3.5. The original plan for building somethin…
This is an outstanding phenomenon, not dystopian. Your friend was going to have to do the incredibly awful and wasteful task of rebuilding an incredibly complicated system. However, fortunately, a much more optimal solution happened, where somebody build a superior and cheap solution (unfortunately not open yet) and let the whole world leverage it. Now your friend needs to waste very little energy deduplicating, and…
Re: What it feels like to work in AI right now
#675From my perspective it’s just _confusing_ to work in AI right now. We have some massive models that are doing some really neat stuff, and apparently hundreds of millions of people are using them—but I keep wondering: to do _what_, exactly? I’m not asking what the models can do, I’m asking what people want the models to do every day, all the time. I’ve been shown some neat pictures people made that they thought were c…
Latvian is a language, that's spoken by less than 2 million people. It's irrelevant outside of Latvia. All of our government funded researchers who worked on natural language processing can now throw their work in the trash and resign. ChatGPT is leagues better than anything that they've done. And OpenAI weren't even trying. Windows wasn't even localised in Latvian properly until very recently. Google translate still…
Real researchers build on advances not quit because of them. I'm sure GPT is not as optimised as it could be to process non-English text. There's clearly a lot of work to do. At least this is true for south asian languages and I'm sure is true even for popular Western languages like French or German
Re: What it feels like to work in AI right now
#676Earlier quoted context omitted.
I've calmed down after your "heads will roll" comment so I promise to be civil. To continue. Yes, you should put back your original comment to reduce confusion. I'm fine reading a long comment. > No idea why you had to tout out that you had a PhD when I was pointing out that you were not a user. You said: And then we have bystanders like you with no understanding of what’s actually going on, who want to ban everybody…
Yea people are going to hate Tesla because of musk or bc they’re always in the news. I’m not saying that there isn’t valid criticism it just seems that they’re held to a totally different standard than other companies. People really enjoy taking a moral stance against them like you’re doing but it really seems like an overreaction.
Nonetheless they released this beta-technology into the public, without consent. It's caused loss of life, property damage, and on top of it all it's also fraud, because it hasn't even delivered on the promise of full self driving after many years of promises.
I don't think calling them out for this is an overreaction.
Re: What it feels like to work in AI right now
#677Earlier quoted context omitted.
You're making incorrect assumptions. This project wasn't about scaling any published approaches. It was original neural net research that produced excellent results with a new architecture without self-attention, using a new optimizer, new regularization and augmentation ideas, sparsity, but with some NLP feature engineering, etc. Scaling it up to GPT-2 size matched its performance for English (my project was English…
Do you have any publications to back up your claims about your work? They seem more than a bit grandiose. If you're ideas are as novel and useful as you say then you should publish them. And I'm sorry, but you're completely wrong about companies recognizing commercial potential. I worked on Alexa for five years, it is a far harder problem than you think. It is nowhere near as simple as "we just weren't looking at the…
If any of the FAANG companies recognized the commercial potential and still accomplished so little, they must be entirely incompetent. When this 2017 deck was created, I had 50k LOC (fewer would be needed now using the frameworks and libraries) plus Word and Chrome plugins. The inference was still too slow and not quite feature-complete, and it was just a writing assistant with several other features in early testing, but it seems more than enough for me to know quite well how difficult is the task.
Re: What it feels like to work in AI right now
#678Earlier quoted context omitted.
You laugh, but this is why SQL reads kinda-sorta like English. People have tried, and failed. Meanwhile, if you give Chat GPT your database schema and ask it to write a SQL query for a report, it can do that for you.
If anyone wants to see the output of GPT4 when asked to define the tables and some sample queries for a hackernews clone in sqlite: https://poe.com/lookaroundyou/1512927999932134
Nonetheless, it could prove useful for looking up algorithms, patterns, and generating boilerplate code. However, an important issue is will it generate similar code if queried at a later time? Not likely, which will make it less useful or result in an inconsistent codebase. Maybe you can request a version of the code generation? In-house code generators will generate consistent code, so it will be interesting to see how it is used in real projects.
Re: What it feels like to work in AI right now
#679From my perspective it’s just _confusing_ to work in AI right now. We have some massive models that are doing some really neat stuff, and apparently hundreds of millions of people are using them—but I keep wondering: to do _what_, exactly? I’m not asking what the models can do, I’m asking what people want the models to do every day, all the time. I’ve been shown some neat pictures people made that they thought were c…
Latvian is a language, that's spoken by less than 2 million people. It's irrelevant outside of Latvia. All of our government funded researchers who worked on natural language processing can now throw their work in the trash and resign. ChatGPT is leagues better than anything that they've done. And OpenAI weren't even trying. Windows wasn't even localised in Latvian properly until very recently. Google translate still…
Re: What it feels like to work in AI right now
#680Earlier quoted context omitted.
Latvian is a language, that's spoken by less than 2 million people. It's irrelevant outside of Latvia. All of our government funded researchers who worked on natural language processing can now throw their work in the trash and resign. ChatGPT is leagues better than anything that they've done. And OpenAI weren't even trying. Windows wasn't even localised in Latvian properly until very recently. Google translate still…
Why is Latvian NLP research in trash? Real researchers build on advances not quit because of them. I'm sure GPT is not as optimised as it could be to process non-English text. There's clearly a lot of work to do. At least this is true for south asian languages and I'm sure is true even for popular Western languages like French or German