AI researcher at Anthropic quits job to write Poetry

[Louanne Learning]

Top AI experts at OpenAI, Anthropic and other AI companies warn of the rising dangers of their technology.

Zoë Hitzig, a former OpenAI researcher, left last week, citing “deep reservations about OpenAI’s strategy” – She writes in a NYT guest essay – “OpenAI seems to have stopped asking the questions I’d joined to help answer.”

Another OpenAI employee, Hieu Pham, wrote on X – “Today, I finally feel the existential threat that AI is posing. When AI becomes overly good and disrupts everything, what will be left for humans to do? And it’s when, not if.”

And - Mrinank Sharma, (PhD in Statistical Machine Learning, University of Oxford) - a top researcher at Anthropic - has quit his job to write poetry.

In his resignation letter – posted below – he writes –

“We appear to be approaching a threshold where our wisdom must grow in equal measures to our capacity to affect the world, lest we face the consequences…

I want to contribute in a way that feels fully in my integrity…

I feel called to writing that addresses and engages fully with the place we find ourselves, and places poetic truth alongside scientific truth as equally valid ways of knowing…

I hope to explore a poetry degree and devote myself to the practice of courageous speech.”


I like his mention of “courage.” I think that is an important human element, or feature, of the creative process.

AI can never have access to the language of the soul and spirituality. AI can never represent with symbols the way humans do.

 

[Stuart Dren]

Quit my job to write poetry.... Why didn't I think of that?

Zoë Hitzig, a former OpenAI researcher, left last week, citing “deep reservations about OpenAI’s strategy” – She writes in a NYT guest essay – “OpenAI seems to have stopped asking the questions I’d joined to help answer.”

It sounds like she wasn't a fan of arithmetic, the same arithmetic that paid her salt in the first place.

Another OpenAI employee, Hieu Pham, wrote on X – “Today, I finally feel the existential threat that AI is posing. When AI becomes overly good and disrupts everything, what will be left for humans to do? And it’s when, not if.”

That take is a bit milquetoast/unimaginative if I'm being honest. If you believe algorithm models can eventually dominate humans at, quote, "everything," (which also implies that robotics will be able to take on any and all physical endeavours, ha, ha) then this dreamer only needs to presume there are future, unimaginable tasks that humans should be doing instead.

Hieu is probably just trying to drum up more publicity for OpenAI, in the same way AI leaders have tried to foment Terminator style panic about the tech in order to keep us from losing interest. They might go public this year, which will be huge.

 

[Louanne Learning]

Quit my job to write poetry.... Why didn't I think of that?

You could always just retire?

 

[Louanne Learning]

Mrinank Sharma,

Today, I downloaded Sharma's book of published poetry - We Live and Die a Thousand Times - and reading this one below - I wondered - is he speaking to his poetry?


My Debt

My Debt Long ago, I made you my debt,
a crying child, who buried, wept.
Today looking back, it is my deepest regret
that I left you back there and tried to forget.

Yes, my dear, I locked you away
afraid of more hurt, that price I did pay.
I tried, my dear, what else can I say?
I left you back there, there to decay.

But today, my love, I went back and saw
that my plans, my acts, they did have their draw,
their hope, a hope, to hide a thought flaw—
you, my dear, so tired and raw,
you, my dear, the source of all awe.

I left on this journey; I thought I must come
and travel to this place in search of freedom,
and to find what I did, I had to become
the child who could sing, the man who could see
that to unshackle those chains, you must be
let be a blossoming petal, so tender and free,
the heart of my heart, a treasure to me.

 

[Naomasa298]

Find him on SM and ask him. He's probably on LinkedIn.

Yep, here he is:

https://www.linkedin.com/in/mrinank-sharma/

 

[Louanne Learning]

Find him on SM and ask him. He's probably on LinkedIn.

Yep, here he is:

https://www.linkedin.com/in/mrinank-sharma/

it was blocked to me unless I join

 

[Naomasa298]

He's got his own website. I'm sure you could ask him there.

 

[Homer Potvin]

That's a terrible poem. Just saying.

 

[Naomasa298]

That's a terrible poem. Just saying.

He should get an AI to critique it.

 

[Stuart Dren]

That's a terrible poem. Just saying.

Well duh, that's why he now has to "explore a poetry degree."

 

[Homer Potvin]

And who puts footnotes in a resignation letter? Something stinks here....

 

[Stuart Dren]

You could always just retire?

I can't find an apparent shred of virtue or even notable character in his act of resignation/letter. Are we talking about him because he's said that he wants to speak courageously? I genuinely don't get it. He made enough money as an AI researcher to be comfortable at a relatively young age in a HCoL area and is now abandoning his primary bread since he has earned the privilege to do so.

It might as well read: "I made enough F-you money to finally F off and possibly become a pseudo-spirituality guru if my website is any indication." Which is of course what I'll write when I retire.

 

[Louanne Learning]

I'm not quite so cynical about the poetic muse

 

[Stuart Dren]

I'm not quite so cynical about the poetic muse

Granted I was being a hater. For the record I don't think his poem is terrible. It's perfectly fine.

 

[Trish]

Granted I was being a hater. For the record I don't think his poem is terrible. It's perfectly fine.

I'm with you. Feels like an insincere publicity stunt to me.

 

[Pyroglyphian]

And who puts footnotes in a resignation letter? Something stinks here....

Academics can’t help themselves.

My take is that this particular academic has strayed too far from the ivory tower and accidentally glimpsed the ugly truth of things. Occupational hazard. Hope he manages to get through the poetry phase of his recovery okay.

 

[Louanne Learning]

Top AI experts at OpenAI, Anthropic and other AI companies warn of the rising dangers of their technology.

Please keep all replies to this in the context of AI, and not politics.

(Mods - if you delete this post, I understand.)

I imagine that the sender of the letter in the OP was involved in talks to have the government use AI in illegal and immoral purposes.

Rattled, he quit.

Now, nearly a thousand employees of Google and OpenAI have signed an open letter standing with Anthropic to refuse the government's insistence that they use their AI models for domestic mass surveillance and autonomously killing people without human oversight.

 

[Homer Potvin]

their AI models for domestic mass surveillance and autonomously killing people without human oversight.

As far as I know, it's a matter of releasing the AI source code to the DOD (or DOW). The mass killings, AI police bots, Skynet and terminators feels a little... theatrical? As a sci-fi writer, I'm glad the tropes can hold as much weight as reality.

 

[Trish]

As far as I know, it's a matter of releasing the AI source code to the DOD (or DOW). The mass killings, AI police bots, Skynet and terminators feels a little... theatrical? As a sci-fi writer, I'm glad the tropes can hold as much weight as reality.

My understanding (which, take with a grain of salt as I didn't verify sources) is that Anthropic refused to sign the new contract unless DOD put it in writing that they would not use their AI for those things because the AI isn't reliable enough for autonomous killing and they don't believe US citizens should be big brothered that hard. So DOD blacklisted Anthropic and OpenAI signed. Now, Open AI claims they got the terms that Anthropic couldn't, but come on. That makes zero sense. Critical thinking says if they were willing to accept that deal and put it in writing they would have stuck with the one they were already using and familiar with. It was literally hours later OpenAI signed their contract.

If that crosses the line into politics feel free to zap it as that wasn't my intention.

 

[Stuart Dren]

As far as I know, it's a matter of releasing the AI source code to the DOD (or DOW). The mass killings, AI police bots, Skynet and terminators feels a little... theatrical? As a sci-fi writer, I'm glad the tropes can hold as much weight as reality.

Independent targeting is a must for drones, given signal jamming. It also makes it scalable, meaning one can overwhelm defences with hundreds or even thousands of drones.

Right now they can navigate independently via comparing topography photos, which probably isn't all that new (Ukraine recently did a complex strike using these to drone an ammo depot).

I'm assuming the intent with the... sigh... DoW is to harness these tools increase independent navigation ability and targeting accuracy.

One argument in favour of the DoW's strategy is that China and friends is going to do it anyway no matter what. Future with autonomous weapons sucks though.

Intelligence is it's own thing. I've no idea. Frankly they would use it either way. It's up to individuals to decide how much surveillance power their govt. should have.

Likewise zap this post if it's getting too political.