Some guy who worked as an artis for Hasbro then started his own game company a year ago and has made one game since isn’t a source of much credibility here.

Companies are actively using AI in their workflows now whether you like it or not, and you can sit and pretend theres some secret cabal of artists (scoff) who somehow are the ones with a say on other artists getting hired (scoff), but thats… not how it is.

The people making the calls are, tbh, incredibly ravenous for anything remotely related to AI. I don’t think its a smart idea, but it doesnt change the fact there is appetite for it in the market, jobs to be had, etc etc.

However, I don’t think a portfolio of AI art is useful… at all

Hiring practices are way more interested in peoples ability to train and refine models, not use them.

I mean using them is sorta useful too, but thats quickly becoming a literal baseline skill that they just expect you are able to do. Prompt engineering is quickly becoming just sort of an expected “you should know how to do this, its not a selling point, its the bar” thing, akin to “you know how to use a mouse and keyboard and write emails” kind of thing.

Not like its hard, prompt engineering is pretty easy to take a 30 min lesson on and be “good enough” so thats why its such a baseline whatever skill.

If you cant do it though (and yes, I have met people who cant figure out how the hell you prompt an AI, they just literally dont get it), you probably will lose your job security very fast. Get used to it lol…

NOTE Im not saying any of the above is a good thing, but it is reality, whether you like it or not.

… Proceeds to use Svelte cuz it has the exact features I want

React is popular but I honestly don’t care about llm weenie vibe coder junior devs being biased towards react. Lions don’t concern themselves with the opinions of sheep.

The majority of shit apps are being made with react sure.

But skilled seniors who intend to make something robust don’t even tend to have an llm enabled that will even influence their opinion on the first place. The majority of senior devs keep repeating the same sentiment: llms primarily are slowing them down more than helping.

Junior devs that crutch on llms are falling behind and the quality of their output shows. They grow slower and produce worse output.

It’s one thing to use it for monotonous tasks, but if it influences your higher level important decisions you are probably already cooked.

What makes that the more likely scenario?

Because it’s their facility

this facility has never had this issue until the FBI showed up to commandeer their incinerator.

Says who?

For all we know they’ve had issues everytime they incinerate but they ignored it cuz a lil bit of smoke from 1 cat is way easier to shrug off compared to a huge amount of meth

It’s very possible they just have been ignoring the problem because normal smoke from incineration a very small cadaver isn’t a big deal, whereas meth fumes are extremely toxic and not something you can just shrug off

Lord knows I’ve worked with workers who have the “I’ve been doing it this way for 10 years and never had an issue, don’t be a pussy” type of attitude too

So hard to say, without more info it’s basically just us speculating.

rather than the FBI for their clear incompetence?

The article has not stated who was responsible for operation of the facility.

It’s more likely the responsibility was on the staff to ensure the equipment at their own facility was functioning right

This sort of error should have been covered by prior operation licensing checks, a facility with an incinerator on premises shouldn’t have negative pressure issues

So something somehow caused a negative pressure issue.

Usually the culprit is some kind of exhaust fan being run, or a door being left open too long

Based on time of year and how hot out it is, I wonder if a staff member left a door propped open or something.

Incinerator systems need positive pressure overall.

Anyone who lives in the north and has a gas based furnace heating system knows how deadly negative air pressure can be…

Read the article:

The incinerator is usually used by animal control officers to dispose of euthanised animals, but local authorities said it can also be used by law enforcement to burn seized narcotics.

The incident was caused when smoke was pushed in the wrong direction because of negative pressure, according to Assistant City Administrator Kevin Iffland.

That sounds like it wasn’t a method specific issue, and if anything had been burnt in that incinerator it would’ve caused the same issue.

Sounds like the facility wasn’t setup right, any facility with an incinerator should definitely have positive pressure, not negative.

Getting a later special meeting request with the ceo, at one company, because he wanted feedback on their interview process itself. He then offered me a different job and I had to decline cuz I already accepted another (this was a few weeks after the initial decline I gave)

In another case they just fast tracked me and I ended up declining the job anyways (didn’t like the job)

I’m full time employed but I still do occasiobal interviews to keep feelers out for how the market is. But I typically decline most offers cuz they’re not good enough to get me to actively quit my current job.

Same, I’ve started just stating I have a maximum of 3 interviews I do before I auto-reject the offer.

This has had interesting results.

In the “right” use case, story points should just represent relative effort.

The hours dont matter, its more about ranking how challenging a task is, in order to help the manager rank the priority of tasks.

You should have typically 2~3 metrics:

1. Points, which represent relative effort of the task to the other tasks you are also ranking.

2. Value, how much value does doing this task provide, how important is it

3. Risk, how risky is it that this might break shit though if you make these changes (IE new features typically are low risk since they just add stuff, but if you have to modify old stuff now your risk goes up)

If you have a good integration testing system automated, Risk can be mostly removed since you can just rely on your testing framework to catch if something is gonna explode.

Then your manager can use a formula with these values to basically rank a priority order for every ticket you now scored, in order to assess what the next thing is that is best to focus on.

I hope you dont use any of the other standard quality of life features day to day that consume substantially more power per day then.

There’s plenty of stuff you likely take for granted every day that you use, that burn way more fossil fuels than training GPT took.

GPT did cost a lot of power, but if you put it beside other fairly standard day-to-day things people tend to take for granted, it’s a drop in the bucket.

• Air conditioning, both at your home if you use it, your work if you have it, stores you visit, etc etc
• Public transit
• Your stove
• Your microwave
• Your water kettle
• Your heating systems everywhere you go
• Your computer
• Your phone
• The internet
• Emergency response systems
• Your clothes washer and dryer

The list goes on and on. ESPECIALLY your clothes dryer, that thing uses a massive amount of power

People seriously underestimate how much power the internet uses overall. GPT’s training provides a concrete, discrete, measured amount of power one specific thing used.

Whereas the internet, as a whole, over one day, uses way more power than all of GPT’s training took total. The issue is “the internet” has its power consumption broadly distributed across the entire globe, in a manner that makes it basically impossible to actually measure how much “total” power you are burning just browsing the web.

But it’s non trivial. Every switch between you and your destination is burning in the range of 150 watts, easily, every router is burning easily 80 watts, etc etc.

And theres dozens of those between you and 1 given destination. The process of routing your packets from your machine all the way across countries at the speed of light, and then a response back, takes a non trivial amount of power. Theres often around 8 to 15 hops between you and the destination, and every single hop tends to have multiple machines involved in that one single packet.

Its easy to handwave that enormous power consumption away because, well, you can’t see it. You aren’t privvy to how much power your ISP burns every day, how much power the nameservers use, etc etc.

GPT is a non trivial chunk of power… but its not THAT much compared to all the other shit going on in the web, its genuinely just a tiny drop in the bucket.

You are extremely naive if you think using GPT makes any kind of notable shift in your total carbon footprint, it doesnt even move the dial at all.

If you actually wanna pick something as a real target for reducing your carbon footprint, the two biggest contenders are:

1. Use public transit, or better yet, bike/walk/run to work. If you work from home, good!
2. Dry your clothes via hangers instead of a dryer

Thats irrelevant to the discussion at hand.

That’s like arguing needles were a bad invention because many people use them for heroin.

People using the tool wrong to hurt themselves doesn’t mean the tool is bad, it just means better regulations and education needs to be put in place.

The training costs effectively enter a “divide by infinity” argument given enough time.

While they continue to train models at this time, eventually you hit a point where a given model can be used in perpetuity.

Costs to train go down, whereas the usability of that model stretches on to effectively infinity.

So you hit a point where you have a one time energy cost to make the model, and an infinite timescale to use it on.

Personally I don’t really trust the LLMs to synthesize disparate sources.

The #1 best use case for LLMs is using them as extremely powerful fuzzy searchers on very large datasets, so stuff like hunting down published papers on topics.

Dont actually use their output as the basis for reasoning, but use it to find the original articles.

For example, as a software dev, I use them often to search for the specific documentation for what I need. I then go look at the actual documentation, but the LLM is exceptionally fast at locating the document itself for me.

Basically, using them as a powerful resource to look up and find resources is key, and was why I was able to find documentation on the symptoms of my pet so fast. It would have taken me ages to find those esoteric published papers on my own, there’s so much to sift through, especially when many papers cover huge amounts of info and what Im looking for is one small piece of info in that one paper.

But with an LLM I can trim down the search space instantly to a way way smaller set, and then go through that by hand. Thousands of papers turn into a couple in a matter of seconds.

The power server side for 5 minutes of chatgpt, vs the power burned browsing the internet to find the info on my own (which would take hours to manually sift through)

Thats the comparison.

Even though server side power consumption to run GPT is very high, its not so high that its more than hours and hours of a laptop usage

Not at tremendously less of a power cost anyways. My laptop draws 35W

5 minutes of GPT is genuinely less power consumption than several hours of my laptop being actively used to do the searching manually. Laptops burn non trivial amounts of power when in use. Anyone who has held a laptop on their lap can attest to the fact they aren’t exactly running cold.

Hell even a whole day of using your mobile phone is non trivial in power consumption, they also use 8~10W or so.

Using GPT for dumb shit is arguably unethical, but only in the sense that baking cookies in the oven is. You gonna go and start yelling at people for making cookies? Cooking up one batch of cookies burns WAAAY more energy than fucking around with GPT. And yet I don’t see people going around bashing people for using their ovens to cook things as a hobby.

There’s no good argument against what I did, by all metrics it genuinely was the ethical choice.

…no that’s not the summarization.

The summarization is:

if you reinforce your model via user feedback, via “likes” or “dislikes” or etc, such that you condition the model towards getting positive user feedback, it will start to lean towards just telling users whatever they want to hear in order to get those precious likes, cuz obviously you trained it to do that

They demo’d in the same paper other examples.

Basically, if you train it on likes, the model becomes duper sycophantic, laying it on super thick…

Which should sound familiar to you.

AI saved my pets life. You won’t convince me it’s 100% all bad and there’s no “right” way to use it.

The way it is trained isnt intellectual theft imo.

It only becomes intellectual theft if it is used to generate something that then competes with and takes away profits from the original creators.

Thus the intellectual theft only kicks in at generation time, but the onus is still on the AI owners for not preventing it

However if I use AI to generate anything that doesn’t “compete” with anyone, then “intellectual theft” doesn’t matter.

For example, when I used it to assist with diagnosing a serious issue my pet was having 2 months ago that was stumping even our vet and it got the answer right, which surprised our vet when we asked them to check a very esoteric possibility (which they dubious checked and then they were shocked to find something there.

They asked us how on earth we managed to guess to check that place of all things, how could we have known. As a result we caught the issue very early when it was easy to treat and saved our pets life

It was a gallbladder infection, and her symptoms had like 20 other more likely causes individually.

But when I punched all her symptoms into GPT, everytime, it asserted if was likely the gallbladder. It had found some papers on other animals and mammals and how gallbladder infections cause that specific combo of symptoms rarely, and encouraged us to check it out.

If you think “intellectual theft” still applies here, despite it being used to save an animals life, then you are the asshole. No one “lost” profit or business to this, no one’s intellectual property was infringed, and consuming the same amount of power it takes to cook 1 pizza in my oven to save my pets life is a pretty damn good trade, in my opinion.

So, yes. I think I used AI ethically there. Fight me.

Exceedingly false representation of the actual experiment.

They took Llama 3 and then trained it further on specific conditions (reinforcing it on “likes” / "thumbs up"s based on positive feedback from a simulated userbase)

And then after that the scientists found the new model (which you can’t really call Llama 3 anymore, it’s been trained further and it’s behavior fundamentally altered) behaved like this when prior informed that the user was easily influenced by the model specifically

What is important to gather though, is the fact that when a model gets trained on the metrics of “likes”, it starts to behave in a manner like this, telling the user whatever they want to hear… Which makes sense, the model is effectively getting trained to min/max positive feedback from users, rather than being trained on being right / correct

But to try and represent this as a “real” chatbot’s behavior is definitely false, this was a model trained by scientists explicitly to test if this behavior happens under extreme conditioning.

If they are mandated, that’s just as bad I agree.

At my company we have tonnes of in house Lunch and Learns (on paid time, non mandatory) that are effectively “I found this super useful thing and want others to know about it”

And I’ll join these things, and see (person), who is on my team, in it too. Later I’ll hat with them about it, or at least try, and they’ll have zero clue wtf I’m talking about.

And it becomes obvious they just joined the meeting to give the illusion of caring, they prolly were afk the whole time. And I suspect this cuz they often do the same for our “in team” mandatory important meetings discussing critical stuff on the project.

Sorta just sounds like you can probably fire a few employees who don’t give a fuck.

From experience, a lot of companies tend to be propped up by like 10% of their developers doing 90% of the work, maybe 50% of developers doing the last 10%, and then like 40% of developers either doing fuck all or actively harming the codebases and making more work for the other 60%.

And more often than not, these people are the ones sending stuff like “AI Note Takers” merely to give the illusion of existing.

In reality you have like three devs who actually care and do most of the work having a discussion and like 10 to 30 AFK participants with their cameras off who probably arent even listening or care.

And the thing is, it shows later. The same devs have zero clue wtf is going on, they have zero clue how to do their job, and they constantly get their code sent back or they plague up some codebase that doesnt have reviewers to catch them.

The AI note takers are just the new version of people showing up to meetings with their camera off and never speaking a word.

Except now they burn orders of magitude more power to do it, which is awful.