Live data from Hacker News

What it feels like to work in AI right now

robotic.substack.com

631–640 of 768 posts

Re: What it feels like to work in AI right now

#631
post #599

Earlier quoted context omitted.

You can't program GTP in anything if you can't program. If your prompt is garbage then the output will be garbage and if you don't know how to program you won't even realize the output was garbage. It's not the language part of programming language that is hard. It's the programming part because it means you have to have a good understanding of what you want. Just like a human programmer won't read your mind an AI pr…

> You can't program GTP in anything if you can't program. That's why it makes this so interesting - this type of automation impacts our jobs directly. Of course, I'm not sure who would use this in a corporate codebase without legal concerns.

Work in a financial place, we have a contract with open AI so they don't hoover our user input. Normal URL is blocked.

Re: What it feels like to work in AI right now

#632
post #550

Earlier quoted context omitted.

GPT-4 is not human level intelligence, nor is is above or below. It’s quite a different kind of intelligence not entirely comparable to humans. That’s probably why we’re moving the AGI goalpost; we visualize AGI as a robot human, but these machines may simply be founded on too different principles to ever receive that honor.

I think it’s mostly different because they crippled the public version for now: no internet access, everything is done in one pass. In our mind we get an idea, we inspect it, we try different variations, we simulate how it will be perceive(consciousness). In this way we iterate before putting the information out. This is not difficult and is getting added on to it externally. Chat GPT is just to get us used to the id…

This is pure speculation, in a very implausible direction.

Re: What it feels like to work in AI right now

#633
post #156

Earlier quoted context omitted.

ChatGPT is cool and novel, but FAANG's requirements for ML/AI go far beyond what ChatGPT provides as a product. ChatGPT is good at answering questions based on an older data set. FAANG typically requires up to date real time inference for huge rapidly changing data sets. Working on the practical side of ML/AI at FAANG, you will probably be working with some combination of feature stores, training platforms, inference…

I can tell you that we have applied teams working on open problems, which can be solved out of the box with ChatGPT. Its a huge deal

What's an example of such a problem?

Re: What it feels like to work in AI right now

#634
post #38
post #26

Earlier quoted context omitted.

It has helped me learn things more interactively because I can ask questions and get answers without having to read 10 blog posts or a book before understanding things. For my personality, it would be hard to recreate without having a living mentor available 24/7 that had knowledge of everything I wanted to ask of it. Does it get things wrong? Sure. But generally the subtle ways it might get stuff wrong would be the…

It's shocking to me how many people are treating chatGPT as a reliable source given how much we know it hallucinates.

It's shocking to me how many people are treating Google Search as a reliable source given how much we know it silos people into extremist sites, and has been extensively gamed by SEO professionals.

Re: What it feels like to work in AI right now

#635
post #279

Earlier quoted context omitted.

I wonder if we saw the same video - or maybe it is just ChatGPT being "great" in the wild? I see one guy asking another guy simple questions and getting weaselwords for an answer.

The interviewer and his questions is in some way more impressive than the answers, which is weird.

Yeah. He's fast and it's just question after question. I didn't realize that was something I wanted to see in interviews.

Re: What it feels like to work in AI right now

#636

Earlier quoted context omitted.

> Let’s go over the basics for Autopilot: Go on Pornhub right now (if you're not at work) and search for sex in tesla. You'll find people driving Teslas on public roads while having sex in the driver's seat. The guy touches the steering wheel every so often to keep the car autopilot activated. The videos have been posted over a span of years. Another one was literally posted yesterday . Combined they have 10s of mill…

No idea why you had to tout out that you had a PhD when I was pointing out that you were not a user. Well news flash, I did my PhD in EECS with a focus on Perception and Autonomous Vehicle Applications from a top 4 CS school and so did most people in Autopilot (Most are from Stanford, Berkeley or CMU in Tesla AP), in fact I worked with people who were a part of the 2007 DARPA Urban Challenge who are in Autopilot now…

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 from using Autopilot because you think drunk people are using Autopilot or something.
Which I took to mean you assumed I was a nonpractitioner who didn't know what I was talking about. I know you work at Tesla, so I know your credentials. Since you seemed so eager to dismiss me as a "bystander", my only point in telling you about my background wasn't to threaten you with it or to assert an unquestionable authority, but to inform you that I have the necessary experience and education to fully understand all the complexities you think are beyond my comprehension. I am not a "bystander", and while I do not own a Tesla (because of course I don't), that doesn't set you up to dismiss my point of view as uninformed. You can communicate to me as a peer, not a "bystander".

> (Most are from Stanford, Berkeley or CMU in Tesla AP)

Absolutely, I've gone to school with some of those people. I've been taught by some of those people. I've also taught some of the people you work with. I don't know who you are, but I know you know better, or at least your colleagues do. Which is why this is so especially painful for me.

> I can assure you they think as deeply about safety as you claim you are.

Do they though? Because... again, you're building a product that allows people to drive around town fucking in their car. I notice you didn't address that at all in your reply. What does your team have to say about that, and when will this be banned? Why did your team release "beta" quality hardware and software onto public streets? Why wasn't the public consulted?

By the tone of your earlier comment about drunk drivers, it seems to imply that you think drunk people driving Teslas is beyond the pale. And yet, what do you say about people fucking in their Teslas? That's happening. Isn't that just as dangerous as drunk driving, if not more so?

It's really easy to say you're thinking deeply about these things, but that seems to be as far as the consideration goes when looking at how your product is being used in reality. You didn't think enough about it that you realized the camera sensors on the AP system would be overwhelmed by a bright white obstruction, and it would cause the AP system to run into it at full speed. You didn't think enough about it to have robust sensing to overcome a single sensor being overwhelmed. Yet you shipped that to the public, and then someone died due to the lack of consideration of that failure mode by your company.

To me, it seems like your company's decisions are based purely on maintaining a competitive edge by being a market leader and aggressively pushing unfinished products onto the general and unsuspecting public. Can you please outline the ethical framework you used to arrive at this decision? Please don't tell me it was "We need this as fast as possible to save as many lives in the future, short term casualties are a necessary evil for the greater good."

> I do in fact sleep soundly, knowing I don't call engineers murderers on online forums.

Thanks for confirming my expectations. I figured as much. I figured you had no problem with the fact that your product decapitated someone, and then your company did literally nothing about it for 3 years, leading to it happening again. For the sake of your sleep, I'm glad you're able to rationalize these decapitations as "it was just a glorified cruise control" as if it's the user's fault. Maybe the first time. But the second one is on your company.

But to be clear I didn't call you personally a murderer; I literally said you're not responsible. But the thing you built is directly responsible. Your team is responsible for releasing these things into communities, which you yourself admit are beta quality. That's something you chose to do. Engineers must be held accountable when the things they engineer hurt people, otherwise they will engineer things that hurt people.

If you haven't realized it yet, this whole feeling I have really about you but your company, so don't take what I'm saying personally, unless in fact you do feel your personal work had contributed to these people's deaths.

Re: What it feels like to work in AI right now

#637

From 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…

To be honest, I can't tell if you are genuinely wondering or just (virtue) signaling that you're not using it, but here's a sample of my own usage: https://news.ycombinator.com/item?id=35299121

Not virtue signaling, but I seem to be asking a different question than people are answering. I’m asking what the product opportunity is and people keep telling me examples of tasks that they use it for.

In many cases the examples are one-off and the only product opportunity is the generative model interface itself. Looking broadly over the replies what I’m seeing is “there’s a thing that used to fail the cost/benefit test, but now the cost is so low that I can automate these things”. So part of my problem is (1) the small benefits of these tasks mean the value proposition comes from volume—that probably comes from the generality of the task engine, and (2) there may be some niche product opportunities on top of the model platform, but the primary big winner here is the platform itself. (That’s not necessarily a new insight, but it seems especially true here.)

The terrifying part is how often I hear people in this thread and elsewhere mentioning tasks that are not fault tolerant to the failure modes of these models. (For example, I had a coworker tell me their relative is a doctor using ChatGPT to diagnose patients.) People keep focusing on the risks of AGI killing us all with paperclips, but I’m much more worried about getting run over by some idiot asking ChatGPT to drive their car.

Re: What it feels like to work in AI right now

#638

From 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…

I have a group of nontechnical friends who use Midjourney constantly for D&D and similar games (both brainstorming and visuals for players)

Re: What it feels like to work in AI right now

#639

AI 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 presumably has free time and energy to think about other problems that aren't solved.

Re: What it feels like to work in AI right now

#640

Earlier quoted context omitted.

> What am I missing? I'm going to guess (assume) you probably haven't worked in a 'real' business. A place where elbow grease still does the majority of the work and where Windows 8 was only just phased out. The killer app (to me) in case of GPT is GPT itself, not ChatGPT. ChatGPT just allows me to easily test use cases for GPT. There are many interesting use cases for those elbow grease businesses for GPT. For examp…

Right, questioning the value of a chatbot is a good indication someone hasn't had real job. Totally legit characterization that doesn't make the whole thing sound like a confidence game.

I don't think donkeyd was trying to insult the person he was replying to, but he definitely could have worded that better. I think he's using "real job" in a derogatory sense directed at businesses, how most of them are so inefficient and rote that they'd stand to gain from the roteness of something like GPT, and that the person he's replying to, perhaps fortunately, hasn't experienced that type of business.
Post reply on HN