Clojurians Log v2

I learned a few things, so I should be able to do this faster in the future. Importantly, this is a generic function and does not require deep understanding of existing code or of my domain. I'm not sure how well it will do in those cases.

Fun exercise, though. A future Meetup idea could be to share LLM generated functions and swap tips.

Well-- that's my Sunday Clojure fun for the day. Time to go for a bike ride. 🚴

It’s funny that you mention this now. I’m prototyping a new application at the moment and trying to use AI as much as I can. Definitely still learning how to interact with it… I’ll look at the code more in a bit! I definitely like the idea for the meetup. I know it would help me a lot to hear how other people are doing these things.

i’ve always used it for more “procedural” goal accomplishing stuff, never for the tricky domain of writing macros. What’s your overall impression? Did it save you time? Was it a good pairing assistant on this?

I'm ok with a meetup focused on garbage generators if you don't mind me playing human advocate the entire time

i think over the next two years it will be very important to know how to use LLMs to increase your output. I very much would enjoy chatting about it

If robots can automate our jobs then what's the point?

Surely the elites will share that abundance that they save by not paying our wages!

And the garbage code they generate, who will maintain it?

And this one macro apparently took a dozen tries to get to. Is that really saving time? It's just using a less specified contextful grammar to ask a non deterministic processor to make text that is likely (but not guaranteed) to be code like text

I’d suggest not committing any garbage. > If robots can automate our jobs then what's the point? I think this cuts a couple different ways. Why waste time on the parts the robots can do quickly? Or if you’re just aiming for drudgery, why not write assembler? Maybe on punch cards?

I don't think software engineering is a waste of time, personally. I enjoy it and like that I can make a living doing it

It's offensive to me to the core that glorified random number generators built on stolen code are being propped up as replacements for real human beings

And I have yet to gain any benefit from them. Every single time I've tried to use one to generate even basic code it leaves something that I'd normally want to throw into the garbage

Then it must have been especially offensive when even simpler machines like the cotton gin replaced humans.

“glorified random number generators” is just not really wrestling with the reality of them and what the future will be. Learning how to use them to augment your own efforts will be more and more necessary i think.

How did the people working cotton gins feel when they lost their livelihood?

I truly hope not. What a bleak future

Probably not great but I’m not sure that means we should have rejected the technology.

Did the people whose lives got destroyed by it benefit greatly from the increased GDP?

I just don't get why so many devs are rallying around a technology that the rich are only interested in to replace them with

These generators exist solely to remove humans from jobs

my view is the future will change no matter what. I can learn how to leverage new tools or wish they didn’t exist.

I prefer to fight for the future where human creativity still exists

And not just manufactured and laundered copyright infringement machines

i’m interested in that i suppose. but how are you fighting?

I argue against usage of golems at every opportunity

I don’t think there exists a future where human creativity goes away so long as humans are alive.

I mean all you have to do is listen to what the AI maximalists say to understand that is the goal

Microsoft just talked about a SWE agent

It is the clear goal to automate our entire profession

I disagree that is possible, but these people are not our friends nor do the AI companies have human dignity and interests in mind. That's obvious as well by the fact that all AI is trained on stolen data

Sure, but you’re simultaneously saying the tech is total garbage and that it is so powerful that it will replace human ingenuity.

I'm saying that right now it is garbage. God willing it will remain so, but the clear goal of the AI folks is total human replacement so why should anyone support that?

"We're announcing a new project to create a total human death bomb. Right now it only works in theory and our tests can only kill a few propel rather than the entire human race."

I think you’re painting with an overly-broad brush there. I would not support anyone who wanted to replace all humans.

It's instructive to follow through what the AI maximalists say to logical conclusions

We have on one end the group that wants to automate all white collar knowledge work

We have another set that want to make physical robots to automate all physical labor

What do you think will happen if one or both are achieved?

Can we look to history to see if, when previous similar automation technologies were developed, the benefits of productivity were equally shared?

Or were the people whose entire lives got upended simply left to rot

Do you honestly believe that the AI overlords will share the benefits with us? There is no universal post scarcity here.

I don’t have time to respond to all of this on the fly. But this clearly seems like a good topic of conversation for a meetup. I think you’re painting another false dichotomy… the benefits of productivity are rarely shared equally. But they are often a win for the average human. Both can be true at once.

I do think we should resist having AI overlords, so that is a point of common ground.

Yeah in that we agree

If this is inevitable and I'm not yet convinced it is, I strongly believe we should rally behind open tech for it

What do you think is the possible path to the technology going away? A global ban? I’m not sure how to think about it not being inevitable.

There have been many AI winters before and hopefully another one comes soon. It's not at all evident to me that they've passed the final hurdle.

Whew. I missed a lot it on e 🚴 ride. I think AI has many more hurdles, but the current hurdle I believe unlocked a lot of economic potential. > What’s your overall impression? Did it save you time? Was it a good pairing assistant on this? Reasonably enjoyable. Not sure if this first attempt saved me time or not. I learned some things about how to use it, that I think will make my future attempts faster. But perhaps the end result is better. In this case, I started down one path, then realized that wasn't the best interface. If I was hand coding it, I may have stopped there-- and maybe dropped in a TODO about how to improve it the future. In this case, I was able to just tell Claude to make the change.

> If robots can automate our jobs then what's the point? This POV may be a little disingenuous. A lot of what we do as software developers is about automating things that will eliminate other jobs. Before there were spreadsheets and accounting software, many humans were maintaining physical sets of book with lists of numbers. Luckily, those people were able to find other jobs. In the end, the increased economic activity led to an expansion in the number of jobs, not a reduction. Here, too, even if Claude created this macro perfectly and bug free, with just a few iterations; I was careful to point out at the top of this thread: > ... this is a generic function and does not require deep understanding of existing code or of my domain Only a very small percent of my workday is tasks like this. Most of my time are much bigger thing about weighing tradeoffs, understanding the impacting of architectural choices and how they relate to my business domain, interpreting requirements, etc, etc. My belief is that these tools can help with the simpler stuff. Leaving me more time to focus on the bigger picture.

I agree that what these tools can successfully automate, if possible, is not all that we do. But you're naive if you don't think that automating our entire profession is the goal of every investor in this technology

> How did the people working cotton gins feel when they lost their livelihood? Sadly, the people replaced by the Cotton Gin were slaves. Sad because that didn't mean an end to slavery, it meant an explosion. The ability to harvest cotton in bulk drove the economy to expand. This led to a much higher demand for slaves to do all the other jobs.

https://apple.news/ADBxsaigxRRSP_FabFMMjrQ

The idea that automation creates more jobs is not always true and historically hasn't been the accepted theory

It's an old debate https://en.wikipedia.org/wiki/Technological_unemployment?wprov=sfti1#History

> But you're naive if you don't think that automating our entire profession is the goal of every investor in this technology I didn't say that I don't believe that. But I think it will take a while and since it's happening, the question for me is: how do we respond? How do we prepare personally and as a society for that inevitable future?

That's fair. I don't agree that it's inevitable though - just that it is the (sometimes unstated) goal of the AI folks. I think the question of how do we respond personally is to hedge bets. I try these things every so often, remain unimpressed, and leave them until I hear some new golden calf has been unveiled. This has repeated for me many times over the past AI craze. If I start being impressed by them I'll try to integrate them where appropriate but that hasn't happened yet. I also try to be the voice of reason in tech circles that tend to be AI positive by bringing the reality of what the goals of the tech is to the conversation

"Yes this thing can help but remember that the goal of the company behind it is your financial ruin" is an important point that is often forgotten or handwaved away

Perhaps the cases where it isn't true are local minima. Some people will be directly impacted. I feel for the taxi drivers in New York who paid over a million dollars for a Taxi medallion, whose value plummeted when Uber hit the scene. I don't think these are untrue. The argument is rather about System level thing: over time, the impacts of technology lead to more opportunity and more wealth across the system. It's whole 'nother debate if wealth equates to happiness, though.

On happiness alone, AI is also a net negative in my experience

AI related tools have lessened my enjoyment of multiple hobbies

So is Social Media. Can you kill that for? As a personal favor.

Photography is plagued by AI fakery. Gaming is attacked on multiple sides: the advent of fake frame generation, up scaling, and so on along with AI demand driving up GPU prices astronomically.

👍 Valid concerns.

Social media has gotten worse with AI fakes plaguing it. Movies have gotten worse as producers attempt to save money by generating frames.

Spotify is now full of fake artists with generated "songs", drowning out real artists

Right this moment it is literally impossible to purchase a GPU above midrange for anything approaching MSRP

Nvidia made about 1000 5090s and they sold out in nanoseconds. They have no incentive to make more since AMD gave up on the high end this cycle and Nvidia can sell the same silicon for 50x the price to corporations aiming to replace humans

That hardware is used to run games that look worse since they're half fakes or more now since it is cheaper for game developers to integrate AI models to fake pixels than it is to pay programmers and artists to make real images

https://venturebeat.com/ai/github-copilot-previews-agent-mode-as-market-for-agentic-ai-coding-tools-accelerates/

As for quality of the graphics and video, I’d say this demo video of Einstein talking and moving is pretty damn good.

https://venturebeat.com/ai/omnihuman-bytedances-new-ai-creates-realistic-videos-from-a-single-photo/

Relevant link I happened across today. https://thebullshitmachines.com/

> For better or for worse, LLMs are here to stay. We all read content that they produce online, most of us interact with LLM chatbots, and many of us use them to produce content of our own. > In a series of five- to ten-minute lessons, we will explain what these machines are, how they work, and how to thrive in a world where they are everywhere. > You will learn when these systems can save you a lot of time and effort. You will learn when they are likely to steer you wrong. And you will discover how to see through the hype to tell the difference. ? I very much agree with this. They aren’t going away, they are useful, learn to use them

From my perusal of it, that intro is decidedly more positive an assessment then the meat of the site

LLMs are extremely good at convincing humans they think, but they do not think

In addition, "they're here, they're useful" is begging the question. These things are essentially brand new and it's not at all proven to me that they're useful. What does it mean to be useful here? Are they useful to the individual? Are they useful to humanity? Are they useful to the rich or the poor alone? I think that's the thrust of the entire debate in this thread. Is it useful to have mountains of code that has been hallucinated by a non deterministic word generator? Is it useful to have less software engineers capable of writing actual code? Is it useful to generate more "art" putting artists out of work? Is it useful to have a huge portion of internet traffic dedicated to stealing content to train these models on? Is it useful to use a fantastic amount of electricity to train and run these models?

Is it useful to give up writing concise, clear, context free code that can be compiled the same way every single time in favor of writing "prompts" to try to get a math model to generate that code?

> “You will learn when these systems can save you a lot of time and effort.” it was your link 🙂

I don't think that's the sole determinant of useful

Not writing tests also saves time and effort initially

Now we can have silicon parrots hallucinate what might be tests to check if the code like words they generated might do what we asked it to do. Is that useful?