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

What it feels like to work in AI right now

robotic.substack.com

Post reply on HN
Print view View original

761–768 of 768 posts

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

#761

Post by cudgy »

int_19h wrote: post #683
cudgy wrote: post #678

Earlier quoted context omitted.

This is a very simple case that doesn’t reflect the complexity of a real project. Like so many attempts before to produce code, using a little effort, it degrades when the complexity level increases even slightly. Once there are more tables, ones that have names which cannot be easily translated from English, it breaks down quickly. These types of tools work ok for brand new projects, but work on existing projects wi…

Here's a more extreme example, using SQL as an API to give model access to game world state to reason about it. https://gist.github.com/int19h/4f5b98bcb9fab124d308efc19e530... Note that in this case it isn't even asked to write specific queries for specific tasks - it's just given one high-level task and the schema to work with. You're right, though, that the effectiveness of this approach depends very much on schema…

> You really need to design the schema around the capabilities of the model for best results, which makes it that much harder to integrate with legacy stuff.

This may end up being a feature of some high level frameworks … “compatible with ChatGPT” or “designed to work with xxx LLM”.

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

#762

Post by int_19h »

cudgy wrote: post #761
int_19h wrote: post #683

Earlier quoted context omitted.

Here's a more extreme example, using SQL as an API to give model access to game world state to reason about it. https://gist.github.com/int19h/4f5b98bcb9fab124d308efc19e530... Note that in this case it isn't even asked to write specific queries for specific tasks - it's just given one high-level task and the schema to work with. You're right, though, that the effectiveness of this approach depends very much on schema…

> You really need to design the schema around the capabilities of the model for best results, which makes it that much harder to integrate with legacy stuff. This may end up being a feature of some high level frameworks … “compatible with ChatGPT” or “designed to work with xxx LLM”.

It will be very amusing if, eventually, our jobs as software engineers will be crafting bespoke AIs to maximize the efficiency of their use by an LLM.

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

#763

Post by JohnFen »

kolinko wrote: post #740
JohnFen wrote: post #713

Earlier quoted context omitted.

That can be increased quite a bit with practice. But it's also not important. It's just the cache memory -- it isn't the limit of what can be learned and recalled.

It is a limit on what you can reason about without a piece of paper. I’m proficient at math, but my working memory is around 6, so I cannot add two three digit numbers to each other in my head (unless I see numbers to be added in front of me).

OK, but I and nearly all of my friends can, so we have duelling anecdotes here.

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

#764

Post by rngname22 »

greeny7373 wrote: post #427
otabdeveloper4 wrote: post #331

Earlier quoted context omitted.

This post is a clever joke, right? Right...?

You never read something on a paper and had no clue how to even start to get it? I talked about snippets of papers to work collueges. Chatgpt can do the same thing. And why wouldn't it? It's only here to help me to find something I can use for further understanding/googling.

For a serious answer, my biggest reservation with using ChatGPT to explain concepts I know very little about is that ChatGPT often just makes up fake information, or very confidently explains things wrong.

Having to fact-check every thing it says feels a bit exhausting.

A bit like talking to a version of Albert Einstein with dementia or alzheimers, who can still say really intelligent things but mixes in subtle bullshit.

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

#765

Post by thomashop »

jazzyjackson wrote: post #653
ThalesX wrote: post #616

Earlier quoted context omitted.

> Perhaps this is the part you're missing. When I've watched people program with ChatGPT it _is_ a lot of back and forth because an enormous amount of context is able to be stored and back referenced. I've tried to build a lot of fun stuff with it so far. Haven't been able to properly 'program it in English' for anything non-trivial. Back and forth ended up in loops of not what I wanted. I'm just utterly confused at…

i've observed this schism between people who can get LLMs to produce useful output and people who are baffled, I think it's a mixture of two things: expectations: using to the LLM to break problems into steps, suggest alternatives, using the LLM to help them think through the problem. I think this is the people using it to write emails - myself included, having a loop to dial in the letter allows me to write the lett…

Interesting point about empathy. Sorry I'm abusing the comment system to get back to your comment in the future.

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

#766

Post by blazespin »

ruskyhacker wrote: post #346
blazespin wrote: post #256

Earlier quoted context omitted.

I dunno, the stock price hasn't really tanked, so it didn't really backfire that much and it allowed them to get a bit of feedback as they try to figure out how to maintain revenue. What if they put out this insanely great model and people just stopped using search? That would be a backfire most likely.

You mean open ai put out a model and people stopped using search? Or Google? What Open ai offers currently can't really compete with search - I understand the data it's being fed gets newer and newer, but it's not really real time like the search engines are. Indexing and presenting data is so different than NLM. Even if it is fed data that's new it's going to have to infer a lot because of a lack of history. It migh…

Well, I stopped using Google personally, except for online docs.

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

#767

Post by blazespin »

arthurcolle wrote: post #756
blazespin wrote: post #254

Earlier quoted context omitted.

Google is in a weird spot. I suspect they are capable of doing so much more, but there is a serious risk to cannibalizing their 99% revenue model (search), which they probably don't yet know if they can monetize in the same way. Unfortunately for them, OpenAI has forced the question down their throat, which I think is exactly what they intended or at least hoped for.

I feel like it's more like that quote/trope "I don't think about you at all" (from OpenAI's perspective, obviously)

Sure, if you understand the underlying point of that quote - if you have to say it, you're probably thinking about the person.

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

#768

Post by wuiheerfoj »

rngname22 wrote: post #764
greeny7373 wrote: post #427

Earlier quoted context omitted.

You never read something on a paper and had no clue how to even start to get it? I talked about snippets of papers to work collueges. Chatgpt can do the same thing. And why wouldn't it? It's only here to help me to find something I can use for further understanding/googling.

For a serious answer, my biggest reservation with using ChatGPT to explain concepts I know very little about is that ChatGPT often just makes up fake information, or very confidently explains things wrong. Having to fact-check every thing it says feels a bit exhausting. A bit like talking to a version of Albert Einstein with dementia or alzheimers, who can still say really intelligent things but mixes in subtle bulls…

Well the alternative is scouring Google for answers and fact-checking them too - ChatGPT cuts out the scouring process
Post reply on HN