Post

(he's allowed to buy mushrooms at the store. he's never ever allowed to come to the forest with my mom and me, fuck knows what he'll sneak into the mushroom basket)

that's why i put most searches into slop evader extension now

Unfortunately, on a very short timeline there will likely be very few traditional search options remaining. Google already seems to be like 50% of what it once was.

right. i can commit accidental tax fraud by myself, don't need to pay anyone to pay claude to make up shit

I use FreeTaxUSA.com! No TurboTax Scammers here!

Perhaps the tax professional will flag sooner what doesn't make sense in Gemini's answer and double check? Perhaps said professional knows which question to ask? Perhaps if anything goes wrong you can blame the professional and its employer (who value their reputation)? I can think of reasons.

The bell tolls for those who don't actually add any value. I'll skip the light fantastic when I can finally stop dealing with accountants.

I say this a lot but really Confirmation bias rules everything around me If you have spent a great deal of time learning about, thinking about, working on that thing... You are going to think differently about that thing than other people do

The number of people who seem not to realize that 1) not everyone majored in computer science and 2) how the median person uses a computer is astounding

I'm a sceptic but due to safety concerns. I'm a social worker who likes how some modalities like dbt can have parts boiled down for wider use. I think a dbt program where the workbook+group+on-call therapist also had a chat app could be good. I'm scared insurance will cut previous things 4 an app

1/3 also an issue The story of Tal Golesworthy shows this really well. He was a process engineer in high pressure systems, and instead of accepting the typical procedure introduced a new one that is very successful Im mostly just agreeing and LLMs are just going to make it all worse LLMs {cont}

Old building structure engineer? I bet thats both awesome and impressively frustrating I cant imagine the number of meetings you must attend Like, at least with a new building, one only has to do tons of corporate meetings I hope you get to do lots of what you actually wanted to do

Indeed. Socmed plays to our social proclivity for gossip. We’re killing ourselves @ the speed of Send. We consumers are entirely too eager to seek out channels that validate our existing opinions & accept (& share/advance) content w/o any concept of source credibility or fact verification.

Well, the MOST bullish seem to be the managers and middle-managers for the obvious reasons.

shout out to this imho _incredible_ @himself.bsky.social thread, which i think super nails the cognito-hazard of over-polarizing oneself. we generally OUGHT work to be receptive to many ways the world is happening around us, even admists clear bad. don't affix thineself. bsky.app/profile/hims...

iiuc the labs are making a net profit on inferrence (or at least on the paid stuff), and the losses are because they're pumping everything into training as fast as they can?

Heh I'm not quite so adversarial as the last half of your statement, especially since I think companies go through lifecycles during which their initial growth tends to produce more social value than they capture and so I wouldnt want to interrupt that - yet But yes it's premature atm

(Tongue planted firmly in cheek) if only there were a trustless decentralized mechanism for assuring service (preferably in TEEs) and a means to do global payments between providers and consumers

of course it's not just about jobs

frankly i’m already impressed by the models i can run on my phone. knowledge is an issue, but their ability to perform purely language-related tasks isn’t really all that far off the big models at all

You can run a decent model on 16 gigs of ram on a M series Mac now. Building agents is a lot like web dev, request response and doing things in between. I have enjoyed working with the smaller ones, it’s interesting what can be done.

one underappreciated fact is that there are like 100 “neoclouds” out there right now that you can go to and run open weights models

I think the cognitive load required to make something that conforms to your firms expectations will still require people regardless of how efficient these models are. Similar to how outsourcing your core product never really works I agree the future is less concentrated tech talent

My most optimistic take is that IT departments in non-software-oriented companies might be able to expand their scope to creating and maintaining inhouse software. Based on the level of expertise still required in the current generation of models, I think that's a reasonable idea

I mean, do we have to wait for capabilities to land somewhere? Seems like there is some monotonic increase wrt to compute scaling. So we can bet "capabilities will keep increasing." If we keep waiting for them to "land" we are going to be waiting till t -> inf. I just work backwards from:

We’ve been interested in a hardware co-op model across the board, including regular compute/cloud stuff. Still in the recurring idle thought phase.

I've had similar thoughts www.newscientist.com/article/2516...

I expect the models will become commodities. Open ones are not that far behind. I see the risk as that the companies de-skill huge numbers of people who become dependent on the convenience of their systems. Then the prices will start going up and the companies will manipulate their users.

that's extremely interesting

Yeah it’s an interesting concept @scoiattolo.mountainherder.xyz

This may become a necessity. Again, if models become "super intelligent" whatever the heck this means and there are only like 3 providers ... that is very very bad under the current conditions. We need to hedge ... but all the talent is concentrated there.

More than just running open models, there is an opportunity to provide small models trained for a specific task. Agents defined for a specific task, do not need a general purpose LLM. They can use a task specific model. This reduces training and inference costs.

i think it's cool that the billionaires get to decide to shake society and see if it breaks and none of us get to do a damn thing about it

I am reasonably skeptical, I merely felt it a useful data point to mention.

Without any understanding you cannot validate outputs

I really don't think one can reasonably argue that the unethically trained machine that makes its users dumber, is driving environmentally destructive industrialization, & is part of a tech industry circle jerk that is almost assuredly going to crash the economy is something anyone should adapt to.

Speaking for myself, I'm mad that every CEO in America is drooling over the idea of replacing humans with a probability distribution in a while loop, which is at 80% as capable as a human on a good day, and dangerously wrong the rest of the time. Which millions of people trust as "intelligent"

I agree that it’s silly to claim that LLMs Can’t Do Anything (obviously they can do many things). I also think it’s silly to claim LLMs Can Do Everything (obviously they can’t). regardless of how one feels about that, this is a very scary time to need a job and people are reacting accordingly

This is the first part where I slightly disagree. Most of us in tech also hate our work. 😂

And it's not just AI taking jobs, it's the unbalancing of the economy, the impact of misapplied AI that will harm profits & therefore job creation, the inevitable bursting of this monstrous tech bubble, the fact that regardless of actual utility powerful idiots will misuse it & damage us further...

My skill set professionally is photography, video production, graphic design, and illustration. I am not in a good place job wise. And I’m 43, so I’m not in a place where I expected to have to retrain for some entirely new career and start back at the bottom.

There's definitely a worry about job security, but there's also a sense of pride in your work – especially for people who see their job as more than "make line go up". We've seen what has happened to journalism as more and more of it has been outsourced to hallucination machines, it's got worse.

I think many critics are also immensely frustrated by how the very real shortcomings of LLMs (shoddy research, misdiagnosis, plagiarism, etc.) are brushed aside because people value the cheap & shiny object over real expertise and accountability

Or even just work that pays a living wage at all hate it or not

The computer isn’t better at trivia though - it just seems confident when it makes up sentences

There’s might also be the consideration that an LLM doesn’t need to be able to do your job, someone just has to THINK it can do your job to cast people aside or put downward pressure on wages.

Many of our customers (in B2B software) have been asking us to make their jobs irrelevant for years (before the current round of LLMs). AFAICT they assume they have the job because of incumbency, but dislike it because it's hard, and would like to keep the former and ditch the latter

All the economic studies I've seen show that economies adapt, not people. When the mill shuts down, the millers don't learn to code. They eat through their savings and fall out of the middle class. More people coming up learn to code, but those workers laid off are just fucked for life.

I mean I have always had very stable jobs 17 years at my last 4 years at current but im hardline against AI and frankly see it as tech that will wreck the world. Use of water, eroding of trust, giving it automation to murder people, stolen works, expected to just consume garbage now, etc

My experience is younger team members have been using coupons (e.g. CVS, Walgreen, Safeway etc.) on a regularly basis Without realizing it because everything's electronic, i.e. you just press more buttons during checkout.

But also now having been in for a few years, I can definitely feel myself starting to lose touch a bit too.

There is no such thing as a "well-intentioned" LLM booster. they are all either morally bankrupt corporate shills or paid grifters.

“It’s hubris!” bro you’re doing a hubris

we’re in a ballyhoo curve

And they never stop to think that this was exactly what the Luddites objected to: having new technology forced upon them that was clearly only going to benefit its owners.

I certainly don't think it's a monolithic group, but I used that shorthand because that is how both writing and Bluesky threads work

There are a few here and there who pop out to call researchers who don't embrace AI stupid every so often, but with assorted honeypots making the less-wary easy to block they are getting few and far between.

Most of them are wildly skeptical of the Kool-Aid crowd, which helps.

This is why I found the weird wagon-circling around this guy perplexing

Periodically I got here and respond to their posts about all the incredible things they are making and engage with them, and after a few back and forths ask them to show me where to buy or download the actual apps they claim to produce and they immediately ghost me 100% of the time lol

I think that this is largely correct.

Yeah, the "fixes" invariably involve humans having to work harder than they would have if they hadn't used AI at all. In addition, AI companies take your inputs & make them accessible to the public, which is why my company forbids putting anything in it we wouldn't mind seeing on a news website.

also the former are somehow a combination of smug and very aggrieved about it

"I'm too valuable. It'll never replace *me*."

idk I have a weird perspective as a tech process person. i've generally had reasonable job security in my career and I am very, very hardcore LLM Skeptic - but also I work in incident management and have seen what happens when shit goes wrong for 20+ years. So...

saying "LLM skeptic" is like saying "child rape skeptic". if you aren't fully grasping the true horror of this technology, you are already on the losing side.

Yeah. I fall into the "have job security in tech" category, and so do most of the people I work with, and I attribute the difference between my attitude toward LLMs and theirs to all the artists I follow on bsky. Without that exposure, I'd probably be an unrepentant booster like my coworkers.

We'd be in recession right now if it weren't for the AI bubble. It's going to pop eventually, and most of the craziness around it will die out with the bubble.

Yes, of course (and I've talked about this a lot in the past)

idk if you're talking to me but i am extremely demonstrably not an AI enthusiast or a fan of these companies or the way they're developing these technologies, on the record

There was an episode of Cautionary Tales about a plane crash caused by a baby pilot trained using AI. The plane got into a Situation, the AI shut down because things were too complicated, the baby pilot was used to the AI watching certain indicators for him, the other pilot wasn't paying attention.

But at least for me in my industry I don't really think I have a choice. I need to learn how to use AI to my advantage because if I don't I won't get a paycheck. I do think it's incredibly powerful. But I also think that most booster discourse is fart sniffing

As someone who is pretty secure and well paid in my tech job, LLM and AI can get absolutely fucked.

As someone who’s only known job security in tech, but seen what American capitalism does to skilled people who do good work when given a chance, I’m anxious I also think there’s something in the air in tech rn, a sense that when it comes to AI you have to get with the program or be left in the dust

I have been working a programming job at the same company for 16 years. I‘m skeptical about not only the underpinnings of the results produced by LLMs, but also the whole rush to build entire business processes around them. Businesses getting way out over their skis to be cheap. houses built on sand

A friend who is an electrical engineer had a situation where another engineer insisted ChatGPT knew details about a circuit board they were using. My friend insisted not because it is an internally developed board, no information on the web. He still insisted it should know. An experienced engineer!

Ultimately, as far as the creation of meaningful or great art/music is concerned, they are just making fancier pencils or paint, or whatever.

some software engineers (like myself) are from a culture/era where copyright is a dirty word, and so are extremely skeptical of people using it as a shield or worrying much about it.