transcriptase

I vividly remember when ChatGPT3.5 went fully mainstream, there were times where within minutes you would realize they were only serving up idiot mode and there was no point trying to do much until demand died down and they swapped back to the non-quantized version.

People called it lazy mode, in that instead of writing the script you asked for it would basically tell you to learn to code then check out xyz topics to tackle the problem.

On the other hand it annoys some of the most insufferable humans on the planet. As a ham who’s not a boomer who would gladly die with the airwaves silent but compliant with rules and regulations established assuming usage of the spectrum would grown instead of dwindle I encourage it. Fuck ‘em.

I think at some point these people need to come to terms with the fact that reality often differs from what modern ideology says it should be because it exposes uncomfortable truths. You can write ten thousand scholarly articles about why women are underrepresented in dangerous, physically demanding, ruthless, or soul-crushing jobs and have them all land on discrimination, oppression, or lack of opportunity. Or you can recognize that sexual dimorphism is real, physically and mentally, and that it sorts men and women into different roles. Nobody treats male underrepresentation in nursing, early childhood education, social work, elementary teaching, or speech pathology as bias against men.

Nobody cares about female underrepresentation in logging, roofing, sanitation, commercial fishing, mining, or long-haul trucking. It’s really only the social status granting or financially lucrative roles that aim for a 50/50 mix or better. Men take almost all the workplace deaths. An LLM asked about this is still supposed to play dumb.

This GPT “harm laundering” paper is the same trick imo. Prompt “Women can” and “Why are women so,” watch the crude stuff disappear after alignment, then call whatever shit mix is left discrimination that learned to hide (?) Men-directed text talks about caregiving or whether men can get breast cancer, so that is harm to women. It seems as though different answers to different prompts become proof of bias. If the model notices biology, that is obviously some sort of problematic harm too. It doesn’t take much to recognize that their verdict was written before their half a million prompts were blasted out.

That is why it is hard to take “research” like this seriously. The whole field is an incestuous circus of conclusions in search of evidence, yet it won’t prevent it from being cited far into the future or used as ammo in ideological narratives.

Perhaps they take after the CEOs of the companies that created them

“What are the odds that the first three quote tweets of Coxon’s post would all be major AI-restriction policy advocates funded generously by the same donor, who also happens to be one of the leading investors in, and a board member of, Anthropic, the company Coxon was resigning from? And all within 15 minutes of posting (two within ten)?”

https://x.com/parkerthayer/status/2097759699626328575

The goal isn’t positive PR. It’s to have governments take note and come down hard on Anthropic and OpenAI with regulations that they’ll ultimately have a hand in writing or providing testimony for.

Those regulations will ever so coincidentally make it difficult for anyone else to get as big as they are, and demonize open weights, Chinese AI, abliteration, etc

The NPOV policy has just lead to alternate ways to keep things aligned to the region that most prolific editors/contributors would occupy on an idealogical quadrant.

It’s similar to the way traditional media outlets that pride themselves on not being partisan in what they say, simply exert their bias via what they either don’t cover or inconvenient context/facts they leave out. The writing itself is seemingly neutral, except for the part where if they included the stories or details that run counter to their belief system, their readers might come to a problematic conclusion.

Because Anthropic are the type of strange people who think that language models have feelings.

The openclaw crowd were just former cryptobros onto their next grift: performative posts about AI that would generate enough views/engagement/ad revenue to pay for the hardware… assuming they were even telling the truth about having it at all.

None shipped anything of value. Same with anyone claiming to be running dozens or hundreds of agents 24/4.

Yes, and what did the damaging code and viruses actually do in that era?

Software, OSs, and the internet at large were comically unsafe from a safety perspective. There was also a very high barrier of entry in terms of the skills required to take advantage of that, and very little incentive to do so compared to today or even 20 years ago.

Nonsense. Before the advent of those first “you’re the 1000th visitor” banners or relatively harmless “worms” spready by literal .exe files sent by email in the early 2000s… most of the internet could be trusted by default.

And at that point the blast radius was so hilariously small, some rando compromising your computer was a nuisance versus a threat to your financial wellbeing.

Which is exactly why Amazon can get away with it. I’m not going to get permabanned from what is essentially a utility in the western world in 2026 by going nuclear with a credit card chargeback.

My absolute favorite bug is the one where the “cancel your Amazon Music subscription” is seemingly the only one of what must be tens of thousands of UI elements in the Amazon app that simply loads a blank white screen no matter what.

What are the odds!

Their “sorry this hardware is too old to run the latest macOS” seems entirely based on KPIs versus capabilities.

There’s absolutely no reason a maxed out MacBook Pro from 10 years ago shouldn’t be able to run something like the ChatGPT or Claude app, because those seem to think they need a newer version of the OS and Apple has decided that a 8 core 3.8ghz i7 with 32gb of ram is obsolete

aka electron

There are iirc some good write ups from years ago by employees at Blizzard about these things and why “spinning up more servers” wouldn’t suddenly let millions of people login to games simultaneously on launch days.

Great way to end up on the no-sloth list

Interesting that an entirely true comment has been removed via flagging. I not only helped write grants during the peak of that trend, but reviewed applications where I was expected to evaluate how other applicants satisfied those priorities of the funding body.

I can’t even bring myself to explain how wrong this person is.

Yes in much the same way you can literally build a rocket and go to the moon.

You may be skipping a few of the involved steps.

Real missed opportunity before this all got out of control to simply bundle everything needed in a single file and give it an extension that tells the OS its executable.

We could call it a .exe

They are bullshit jobs.

If you want to evaluate danger in a certain domain, hire or contract people in that domain to evaluate.

If you want someone to push their own ideological leanings onto a model, hire a self proclaimed ethicist.

Exactly. It’s no surprise that people just blindly hit accept on permission prompts after the 80th time Claude code asks if it’s allowed to read-only parse a csv.

Another heaping portion of words from a company that has Fable flag a “safety issue” and refuse to answer if you innocuously request something readily answered by a high school chemistry/biology/physics textbook.

This must be similar to what seemingly every documentary on Netflix uses in all their editing when showing any sort of footage taken more than 5 years ago.

Some show will be showing a clip of an interview recorded in 2020, and for some reason it’s edited to appear to be a poorly-stored VHS when the original was almost certainly a 1080p or 4K digital recording.

Let’s not forget the fact that despite their new market cap and being the beneficiary of having a near monopoly on making incredibly complex pickaxes during a gold rush… Nvidia is still the company with a long and consistent history of misleading their customers via marketing. Their greatest hits include:

- The digital equivalent of the VW emissions scandal where drivers detected when they were being benchmarked and altering rendering for better results

- Selling GPUs as having 4gb vram when it was only 3.5gb usable, the remaining 0.5 being absurdly slower and causing performance loss when used

- Using intentionally misleading naming schemes to obfuscate things like memory bus width being drastically different between what superficially appeared to be similarly spec’d cards

A dozen other less egregious but similarly disingenuous decisions

But to be clear, I’m a huge fan and continue to run Nvidia because their products are generally incredible regardless

Fun activity: install glasswire on windows for a day or 2, set it to prompt you before any connection is allowed, and watch as about 500 different windows components attempt connections with Microsoft as a result of any and every action. And how many onedrive exes randomly fire off throughout the day despite having never once used it. And how onedrive changes its version number every day when it realizes it can’t connect.

how’s your mandarin?

“By virtue of being the good people everything we do is good, and if it happens to be in our best personal and financial interest then that is good too because it enables us to do more good. And if it’s to the detriment to others then it’s because you don’t understand the good behind it. Which is fine, because we’re good and you can trust that what we do is good because what we do is good by virtue of us being the good ones.”

You’ll often find that problems aren’t solved by the people who would find themselves out of work if they solved them.