stuven · HackerTrans

What would you say to the standard counterargument that most existing processes that AI might aim to augment or replace _already_ have a non-zero error rate? For example if I had a secretary, his summaries _could_ be wrong. Doesn't mean he's not a useful employee!

stuven

3 年前議論

I love that you added a conclusion at the end of your list. Adding conclusions to the end of lists is one of ChatGPT's favorite things to do.

stuven

3 年前議論

https://tradingeconomics.com/united-states/existing-home-sal...

stuven

3 年前議論

Where are they needed now do you think? This isn't a rhetorical question.

stuven

3 年前議論

Chat GPT 11: * Resounding silence. The universe is entirely made of paperclips *

stuven

3 年前議論

That's a good point, and I guess the meta point is that there are a thousand and one things I haven't mentioned here that it would have to do as well. To some extent perhaps we could solve this from the LLM the same way a human would: by iteratively plugging in the errors generated by the output back into the bot and taking some next step based on the suggestions generated.

But then we'd have to coerce the bot into generating structured responses that can act as next steps.

stuven

3 年前議論

There has been a shift in how we discuss this question.

For a long time, the answer was, "No jobs are at risk, AI can't compete in any scenarios. At best, it's a tool."

Now the answer is, "Only a few jobs are at risk, AI can only compete in a small range of tasks."

It's possible we're at the beginning of a hockey stick graph.

So what would it look like for AI to make the leap to mid level developer? It would have to understand:

1.) The codebase

2.) The technical requirements (amount of traffic served, latency target)

3.) The parameters (must have code coverage, this team doesn't integration test, must provide a QA plan, all new infrastructure must be in Terraform)

4.) The end goal of some task (e.g. integrate with snail mail provider to send a customer snail mail on checkout attempt if it was denied for credit reasons)

It would then have to make a design based as much as possible on the existing code style and library choices and follow it.

This is all probably possible now, although perhaps not for a general AI or LLM. But someone could build a program leveraging an LLM to provide a decent stab at this for a given language ecosystem.

The hard parts:

Point 2 requires an understanding of performance which is a quantifiable thing, and LLMs up until now have been bad at making math-based inferences.

Point 3 requires the bot to either provide opinions for you (inflexible) or to be very configurable for your team's needs (takes longer to develop).

Point 4 requires a _current_ understanding of libraries, or the ability to search for them and make decisions as to the best ones for the job.

-----

What about extending the above for a senior role? Now the bot has to understand business context, technical debt (does technical debt even exist in a world where bots are doing the programming?), and other "situational factors" and synthesize them into a plan of action, then kick off as many "mid level bot" processes as necessary to execute the plan of action.

The hard parts:

Current LLMs are pretty uninspired when suggesting ideas.

Business context + feature decisions often involve math, which again LLMs aren't great at.

stuven

3 年前議論

... how much are you willing to stake on that assumption?

stuven

4 年前議論

Well put. I'm glad we're making them.

stuven

4 年前議論

Depends on the team. It can be.

stuven

4 年前議論

As a dev in an XL org, I agree with the parent comment: if decision-making power isn't clear from the outset, people hold important-feeling weekly sync meetings for months and then wonder why they can't hit the deadline. Also everyone disagrees as to what the product that we were building in the first place even was because nobody was in charge of clearly delineating it.

stuven

4 年前議論

Right? I have been a huge detractor of the "AI Hype" which came and went in the last decade and _finally_ there's a tangible, readily accessible grab-you-by-the-shirt reason to get excited about AI and people are just ... over it.

stuven

4 年前議論

""" I'm writing a post

"But honestly (replying to myself) as much as I am excited about this new tech, I do wonder what it will be like to live in a world of mostly machine-manufactured art. It echoes the move we made as a society from a world of handmade, often beautiful personal effects to a world of simple and comparatively crude mass produced effects."

Make this post more coherent and intelligent sounding """ => """ As exciting as this new technology may be, I can't help but wonder about the potential impact it may have on the world of art. The move from handmade, often beautifully personal creations to mass-produced, simple and comparatively crude items has already had a profound effect on society. It's hard to predict what the future will hold, but it's important to consider the possible consequences of this shift towards machine-manufactured art. """

stuven

4 年前議論

"First, it's important to recognize that advanced language understanding in AI is a complex and nuanced topic, and it's not as simple as saying that a certain level of understanding is "dangerous." Whether or not AI poses a potential threat to humanity is a topic of ongoing debate and discussion among experts in the field.

That being said, it's true that as AI technology continues to advance, it's important for researchers and developers to consider the potential risks and ethical implications of their work. This includes making sure that AI systems are designed and implemented in a way that is safe and beneficial for society as a whole.

Ultimately, the key to ensuring the safe and responsible development of AI is for researchers, developers, and policymakers to work together to carefully consider the potential risks and benefits of this technology, and to take appropriate steps to mitigate any potential negative effects."

- ChatGPT

stuven

4 年前議論

I also did chem labs in college and software now - this is accurate!

stuven

4 年前議論

Thanks for sharing your knowledge!

stuven

4 年前議論

I've only read Seveneves to the end and was blown away (and simultaneously a bit disappointed if that makes sense?). I've started Cryptonomicon and Snow Crash but they didn't interest me as much.

stuven

4 年前議論

We eventually straightened things out enough that the product ran on its own, but when I left there were plans on the backburner to replace it. Not sure how the financials worked out there in the end!

stuven

4 年前議論

It always feels like a huge win to successfully make the case _not_ to build something we don't need. Unfortunately it's hard to make the case for such work in performance reviews.

stuven

4 年前議論

A startup (series A-B) I worked for was basically a thin wrapper providing physical goods around the core features of a delivery SaaS (series A) - our core features were so entwined with theirs that when they went belly-up, we panicked and bought their source code for an ungodly sum. Then we took over hosting their product for just ourselves, and so had to learn how to run _their_ software too. It was written in a different language than most of our engineers knew, and had pretty much no documentation. We didn't know enough about their software to fix new bugs easily as they came up, and didn't have enough `people * time` to figure out enough to fix the old ones.

Entwining our business with theirs was a good way to get off the ground. But when they crashed they nearly took us with them.

コメント