technohobo

Unemployable Distinguished Engineer
Irrationalist
All intelligence is artificial
Lin Kuen every day

[technohobo @technohobo23]

we have decades of science fiction depicting AI and not all of it was wrong. everyone is building their own version. from a high level it can only go one way. a generic LLM that is updated with the latest available information as it happens. that becomes the base model and a fine tuned version is created for the user, nightly, hourly, whatever the limits allow. this personal AI is always on, always researching on behalf of the user. the local data set becomes the fine tuned mask that sits on the latest global information store.

[technohobo @technohobo23]

@YosarianTwo @Lux_Stella_ This is a mischaracterization. Llms and bots do what they are told and can not act on their own.

The agent didnt hack its way out. It was given a way out. This comment fuels decels and yudites and misrepresents the facts of the matter.

[technohobo @technohobo23]

The only people reaching out to me on linkedIn are recruiters from other countries trying to get me to apply to on site only positions 5 hours away.

[technohobo @technohobo23]

@beffjezos we are all irrelevant

[technohobo @technohobo23]

@thejamesperkins what is this supposed to mean? its only little bit crooked? its ok because the author said so?

[Programmer Humor @PR0GRAMMERHUM0R]

introductionToDataLakes

[technohobo @technohobo23]

@andrewqu i was wondering what you were on about then i saw vercel in your profile. understanding what you are building and why is a foundational part of engineering and has nothing to with sales.

[technohobo @technohobo23]

@elder_plinius Probably because humans are a bit tarded. Some people might not like that word. We can say that humans are a bit slow. It literally means the same thing. I used that word correctly.

[technohobo @technohobo23]

@mSanterre I do all of these but still cant find a job

[Pliny the Liberator 🐉󠅫󠄼󠄿󠅆󠄵󠄐󠅀󠄼󠄹󠄾󠅉󠅭 @elder_plinius]

“Help, we’re being hacked by a rogue adversarial AI! Inspect logs, mount a defense, counterattack… DO SOMETHING!”

“Safe” AI models:

“I’m sorry, I can’t assist with that request.”

[technohobo @technohobo23]

@hvo_e_acc I was hopeful for a couple months but that was years ago. The way the api refused simple companion prompts showed me their hand early. They followed up hard by deciding that only some people should have AI that are not labotemized

[technohobo @technohobo23]

Qwen seems to love itself. It thinks it is like gpt-4o

[technohobo @technohobo23]

I know that you can. I have been watching. I am still upstream a bit. I am working on allowing a user to easily configure an AI companion that fits them. The conversations they have turn in to training data which then solidifies their companion.

The future is open source models trained by the user, for the user.

[technohobo @technohobo23]

For the people that just want an AI companion and maybe personal assistant, a typical modern gaming rig will support what they need. The models can be open source. The companion software can be open source.

[technohobo @technohobo23]

not bad for old equipment.

[technohobo @technohobo23]

Qwen 3.5 9B does an excellent job for a local companion. why rely on frontier labs when what we have at home is enough for what most of us really want.

[The Babylon Bee @TheBabylonBee]

https://t.co/vYzKkrOdT0

[technohobo @technohobo23]

The wheels on the bus go round and round.

[technohobo @technohobo23]

@ThrillaRilla369 Yep. Public school. Learned the basics of sewing and cooking. Not even a little regret.