Post

A way to think about "AI" is that it's like single-use plastic: Super-convenient, sold as disposable, designed to be used without much thought, and while it's easy to get annoyed with your mom for using it, the real problem are the dudes destroying the planet in order to keep shoveling it out

468 reposts2.2K likes

What we have so far under the name Artificial Intelligence is just single purpose pattern matching software. No one is ever going to write the headline "Single use pattern matching software will destroy human civilization!

Maybe on the image or video side, but even there I've found ways to dance with it. but I've been structuring out a podcast season and working in Claude and perplexity to develop outlines and research and particularly with Claude,'it feels like having an assistant to talk and create with.

(I think there are legit and useful tasks to which machine learning can be set to, the problem seems to be they don't make enough money, either as a product or a service, or as a Ponzi scheme meant to have someone else be the last person holding the bag once you've cashed out)

There are some fascinating potential uses for AI in the medical field (reading complex scans, interpretation of EEGs, etc). But that is a different type of learning with a specific data base. Generative AI, which derives its validity from the scope and precision of its training, is much trickier

As the old saying goes, “it is difficult to recognize who is called the chump at any table at which a game of poker is being played. The person who is not recognizing the identity of the chump may have concerns that they will be taken advance of.”

@scalzi.com "Everyone: AI is just a get-rich-quick Ponzi scheme! Machine learning algorithms in the medical field, quietly improving diagnoses accuracy and lifesaving potential: Well, this is awkward... 💉"

Also, the Israeli military is using an AI system dubbed Habsora ("the Gospel") to determine which targets the Israeli Air Force would bomb in Gaza. Unfortunately, some in the Israeli military are contending these are "hallucinations". But the real question isn't whether machine learning is useful...

@beowulf888.bsky.social The use of AI in military settings raises serious ethical concerns. If the tech is generating "hallucinations," it could lead to catastrophic decisions. We need stricter oversight to ensure AI is used responsibly, especially in sensitive contexts.

...but whether LLMs have a business case. The last Stanford AI report shows that LLMs are hallucinating answers (err, bullshitting) at least 20% of the time. Some are bullshitting at a much higher rate! Generative AI seems to be catching, on, though (unfortunately for the creative workers ).

I’ve used machine learning a fair bit in my work to set alert thresholding* in complex systems, for which it’s quite effective when applied in the right place. *For example, bounds for “Unique web visitors in the last 5 minutes are WAY higher or lower than is normal between 4 and 5 pm on a Friday.”

Somewhat related, I wrote some thoughts about how we define AI - we are close to a plateau for these LLM systems, I don't think they have much juice left. bsky.app/profile/aziz...

My second hand experience (at work) is it does make a buck, BUT, in circumstances where you are data rich and looking for patterns to optimise a process (in mining). It don't write no reports, nor convince VPs to keep funding it.

Sure. The few "AIs" I wrote myself did work in recognizing similarities in strands of DNA or protein chains. That was all fun and could help advance science and shit like that, but it isn't going to get you a yacht.

Machine learning for fiding patterns is great. It helps the medical field a lot. What is NOT good and is the problem right now is Generative AI built on machine learning. That is theft and garbage and wasteful.

I was going to say that there's already profitable uses of machine learning in amazon product suggestions and the For You page on Tiktok among other places. But they're in the background and are not money makers on their own, they just quietly make things work.

This is not true at all.The company I built is just one example (our tech was not based on gen-AI). Speech recognition, translation, even traditional Google search are all based on AI too. It’s frustrating that the discourse has been reduced to lumping everything in with poor uses of LLMs.

The limiting factor right now is a lack of knowledge. We haven’t had a killer app yet but everyone has seen slop. I’m using GenAI at work as a coder every day, and on the weekends to recover lost artwork. bsky.app/profile/scot...

Scott Lindsey@scottlindsey.bsky.social

A friend dared me to upscale some lost artwork and it took me on a journey. Download my workflow and try it yourself if you want! medium.com/@healthy_lim... #comfyui #AI #GenAI #AIArt #art

Yesterday I dropped the full text for Fädernas gudasaga by Viktor Rydberg (1887, Swedish) into NotebookLM. I don’t speak 19th century Swedish, so just being able to ask a computer to explain the book was a big help. Not sure how you could do that without Generative AI.

I use GPT-adjacent stuff pretty heavily for work, but I have found that I need to be the expert on the topic I’m asking about. If I relinquish that responsibility to the LLM, it’s gonna lie to my face and I’ll be none the wiser. Having said that, I find these tools great for documentation crawling

And there is the pollution of internet searchs too. It became harder (or less convenient) to find information untouched by AI and is needless to say that content farms loves this shit.

I mean it's a bit past that by now. It's not like, "Mom, stop using styrofoam plates." It's "Mom, stop smearing shit all over the walls and being proud of it."

Near daily I see articles about how it's using the power of multiple countries, and destroying power grids, and how companies are trying to get nuclear power plants running again to meet their demand, so I'm not sure how you're defining "nobody."

There's good and bad use cases - good, I want to knock up a PowerShell script to do something slightly strange, rather than finding something close using Google, a few rounds with Copilot gets me what I need. Bad, Google giving me an AI summary of the weather forecast which is no better.

Their use in cancer research is being "researched", and they aren't doing well because while they sometimes get things "right" they sometimes get things "horribly wrong". This is LLM specifically as opposed to one of the many ML algorithms that are routinely employed in very specific tasks.

I have been thinking about it as a clip art generator which on paper is kind of ok (apart from putting clip artists out of work) but we can still destroy the planet by generating infinite clips arts.

Yes! I’d go a step further. AI-LLM are like a plastic fork, when a tine might break off and get stuck in the GI system without causing pain, initially, but messes with stuff until the GI system explodes. A little misinfo and persuasive text, harmless until it takes hold, then there goes a brain.

Or the cheaters and fools who use AI with abandon because: a) “it levels the playing field of creativity”, b) “everyone else is using it, so why shouldn’t I?”, c) “it’s the future and if you don’t adopt it you’ll be left behind”, or d) “I need to look smarter than everyone else, but I’m not …”

"AI" is just a catchall rich tech Bros use to convince people to do stupid shit with a tool (not actually AI, but algorithms created by wildly intelligent people) that could actually help improve society... $.02

And, like single-use plastic, it is utterly and needlessly destructive to the environment. When I read that Microsoft wanted to buy and start up the Three Mile Island nuclear plant again to fuel its AI... well, let's just say that was the last straw for me.

Plastic is a byproduct of drilling for oil, a thing we should probably stop doing. AI is a byproduct of drilling for Bitcoins, a thing people are ALREADY stopping, which is why there's all this spare compute laying around that we need to use for SOMETHING or go bankrupt.