f'n Ai

[Coss]

And logic, that neither Ai, nor us clever folk, can fathom

Used to be, when talking tech, you could say "cut the cord".  I now take this to be "turn off electric"

Ai has some way to go, before it's biologic.

_____

flat cup, chick, slug, slug, bird, bird.

𓎢𓅱𓆑𓆑𓅂𓅂

coffee

[Coss]

and in the Ai safety arena:

Anthropic CEO Dario Amodei is meeting with Defense Secretary Pete Hegseth today, as the Pentagon threatens the AI company with what could amount to a government blacklist.

At issue is the guardrails Anthropic placed on its AI models. The Pentagon, which has a $200 million contract with Anthropic, wants the company to lift its restrictions for the military to be able to use the model for “all lawful use,” according to a source familiar with the discussions.

But Anthropic has concerns over two issues that it isn’t willing to drop, the source said: AI-controlled weapons and mass domestic surveillance of American citizens. According to a source familiar, Anthropic believes AI is not reliable enough to operate weapons and no laws or regulations yet that cover how AI could be used in mass surveillance.

LINK

_____

Seems to amount to,  US Defense Secretary Pete Hegseth, shouting "Gimme, gimme, gimme, gimme...",  and Anthropic Co-founder and CEO Dario Amodei, trying to get Hegseth to understand the dangers involved.

[Coss]

I was just listening to a guy (apparently a professor at Oxford) explain current Ai to us mortals.

His message is that Ai large language models, are guessing based on popularity. He related a test that one of his colleagues did which was to ask "How many r s in the word Strawberry" the Ai model confidently said there were two.

The point of this is that it did not go and count the letters,  but it saw that of all the data it had access to, that the majority of folk thought there were two. And on this basis it guessed that 2 was the right answer.

Ai has learnt about this now because of this example becoming more very popular.

Again (I've posted about this before)  Popularity ≠ Truth

 

 

[Coss]

And on the question I asked  Ai  "Which is more true, Science or Religion?" Ai wants to play both sides.

I can feel a crisis coming on, in the medium term future.

In our future that is the movie "Idiocracy",  the computer (Ai),  will become the "Yes" computer, rather than the "Computer say No." that it is now.

[Coss]

"A system designed to predict the most likely next word in a sentence can also write good computer code, offer strategic advice and respond with remarkable empathy to human problems. We don’t fully understand why. And yet the dominant instinct across the corporate world is to treat artificial intelligence as if it were just another piece of enterprise software: slot it into existing processes, assign IT Key Performance Indicators and hand it to the it department for management.

This is a profound strategic mistake. Companies are racing to de-weird ai, and in doing so they are squandering what makes it transformative, turning it into just the latest wave of office automation. To be clear, making ai easy to use is important, and building it into familiar tools and workflows is just smart engineering. The mistake is letting those smooth interfaces flatten your understanding of the technology’s possibilities..."

Ethan Mollick - The Economist

LINK