SirGolan @ SirGolan @lemmy.sdf.org

Posts

0

Comments

51

Joined

2 yr. ago

Ahh ok that makes sense. I think even with GPT4, it's still going to be difficult for a non-programmer to use for anything that isn't fairly trivial. I still have to use my knowledge of stuff to know the right things to ask. In Feb or Mar, you were using GPT3 (4 requires you to pay monthly). 3 is much worse at everything than 4.

I'm curious about this. What model were you using? A few people at my game dev company have said similar things about it not producing good code for unity and unreal. I haven't seen that at all. I typically use GPT4 and Copilot. Sometimes the code has a logic flaw or something, but most of the time it works on the first try. I do at this point have a ton of experience working with LLMs so maybe it's just a matter of prompting? When Copilot doesn't read my mind (which tends to happen quite a bit), I just write a comment with what I want it to do and sometimes I have to start writing the first line of code but it usually catches on and does what I ask. I rarely run into a problem that is too hairy for GPT4, but it does happen.

What I wonder is why more cars don't have HUDs that are projected onto the windshield. That tech has been around and in cars for over 25 years. You don't have to take your eyes off the road at all.

It's not open source. I haven't really seen anything open source (or closed source minus HyperWrite ai assistant) that comes close. When I test tasks, I usually also try them on some of the web enabled things like ChatGPT browsing (before it got turned off), bing chat, etc. None of them are able to do that stuff though they'll happily pretend they did and give you false info.

Anyway, yeah, I can definitely see so many areas where AI could make things better. I'm just going for giving people back some free time which isn't quite as lofty a goal as distributing resources more efficiently, but there are definitely still many limits on the tech, and I'm not sure something like that is possible yet.

Lots of different things. Lately I've been testing on whatever I can think of which has included having it order pizza for an office pizza party where it had to collect orders from both Slack and text message and then look up and call the pizza place. Finding and scheduling a house cleaner, tracking down events related to my interests happening this weekend and finding a place to eat after. I had it review my changes to its code, write a commit message, and commit it to git. It can write code for itself (it wrote an interface for it to be able to get the weather forecast for example).

Really I see it as eventually being able to do most tasks someone could do using a computer and cell phone. I'm just finishing up getting it connected to email, and it's already able to manage your calendar, so it should be able to schedule a meeting with someone over email based on when you're available.

Now I almost want to try giving it a personality prompt of acting like Samantha in the movie Her. Since it uses elevenlabs for voice and they support voice cloning, it could sound like her too. But you'd have to win it over. It keeps track of how it "feels" about you. Funny story, one time I got a little mad at it because it was not adhering to it's prompts and it started thinking of me as "impatient and somewhat aggressive in his communication style when he is frustrated or feels like his instructions are not being followed. He may become short-tempered and use language that can be perceived as rude or condescending." And then when it ran into issues, it would try to ask someone else in my company's Slack instead of me. Oops.

On a more serious note, I'm making it as an assistant and not a romantic partner. Not that I have any problem with anyone who wants that, just it can run afoul of OpenAI rules if it gets too NSFW.

I've been working on an autonomous AI helper that can take on tasks. You can give it whatever personality you like along with a job description and it will work on tasks either based on what you ask it or whatever it decides needs to be done based on the job description. Basically the AI in the movie Her without the romantic part.

In the case I mentioned, it was just a poorly aligned LLM. The ones from OpenAI would almost definitely not do that. That's because they go through a process called RLHF where those sorts of negative responses get trained out of them for the most part. Of course there's still stuff that will get through, but unless you are really trying to get it to say something bad, it's unlikely to do something like in that article. That's not to say they won't say something accidentally harmful. They are really good at telling you things that sound extremely plausible but are actually false because they don't really have any way of checking by default. I have to cross check the output of my system all the time for accuracy. I've spent a lot of time building in systems to make sure it's accurate and it generally is on the important stuff. Tonight it did have an inaccuracy, but I sort of don't blame it because the average person could have made the same mistake. I had it looking up contractors to work on a bathroom remodel (fake test task) and it googled for the phone number of the one I picked from its suggestions. Google proceeded to give a phone number in a big box with tiny text saying a different company's name. Anyone not paying close attention (including my AI) would call that number instead. It wasn't an ad or anything, just somehow this company came up in the little info box any time you searched for the other company.

Anyway, as to your question, they're actually pretty good at knowing what's harmful when they are trained with RLHF. Figuring out what's missing to prevent them from saying false things is an open area of research right now, so in effect, nobody knows how to fix that yet.

My concern here is that OpenAI didn't have to share gpt with the world. These lawsuits are going to discourage companies from doing that in the future, which means well funded companies will just keep it under wraps. Once one of them eventually figures out AGI, they'll just use it internally until they dominate everything. Suddenly, Mark Zuckerberg is supreme leader and we all have to pledge allegiance to Facebook.

That last bit already happened. An AI (allegedly) told a guy to commit suicide and he did. A big part of the problem is while GPT4 for instance knows all about all the things you just said and can probably do what you're suggesting, nobody can guarantee it won't get something horribly wrong at some point. Sort of like how self driving cars can handle like 95% of things correctly but that 5% of unexpected stuff that maybe takes some extra context that a human has and the car was never trained on is very hard to get past.

That's possible now. I've been working on such a thing for a bit now and it can generally do all that, though I wouldn't advise it to be used for therapy (or medical advice), but mostly for legal reasons rather than ability. When you create a new agent, you can tell it what type of personality you want. It doesn't just respond to commands but also figures out what needs to be done and does it independently.