Anthropic Warns: Slow Down

Anthropic founders penned a blog post on Thursday, warning that while AI has not been able to self-generate autonomous capability, this is not inevitable.

The blog is asking private industry to put the cart before the horse and solve the societal and existential issues, prior to releasing software.

David Sacks, the lawyer who has personal investments in companies like Facebook, Palantir, Crypto, Paypal, etc. and is the AI and Crypto czar for Trump, says bullshit, and accuses Anthropic of hyping stuff for marketing purposes and of wanting to control things through regulation.

Trump is just as clueless as he is about everything else, but Sacks' portfolio will get fatter if we do not heed Anthropic's warning.

As an Atheist, I'm rooting for the Pope, who had the co-founder of Anthropic in the audience in the Vatican when he issued his "slow down before you cause an ELE (extinction level event) " warning.

Top | New | Old

DeWayfarer · 61-69, M

I agree with slowing it down a bit. Yet for totally different reasons. Some decent laws do need to be made.

Just that alone is a justifiable reason.

Ecological reasons because of heat from the data centers as well are a justifiable reasons.

Economic reasons as well! Where is the electricity going to come from? 🤷🏻‍♂

I really don't see AI changing it's own training data or the limiting factors already in place. Which what anthropic is referring to.

The hardware just isn't around, even with quantum Chips. If you haven't noticed the price increases in memory storage you should! There's not enough. AI certainly can't create it's own hardware.

Recursion in programming AIs is a memory problem! Exceed the limits it will most certainly crash hard!

SunshineGirl · 36-40, F

Most folk seem to agree that strong political leadership and regulatory oversight are essential . . up to the point that a principled politician actually seeks to place the public interest above the unrestricted pursuit of private gain. Then we all remember that the state is a bureaucratic dinosaur that should not be allowed anywhere near AI (apatt from bunging the industry massive subsidues and public contracts, of course) 😩

Ontheroad · M

I am somewhere in the middle. With the uber wealthy and powerful involved and throwing in the ignorant (like Trump), I see both what the Pope had to say, and what Anthropic has to say.

View 1 more replies »

Ontheroad · M

@Northwest I misread, or actually skimmed over it. Still I'm somewhere in the middle and want AI to move forward with all the possibilities it gives us, but I do think we need to think deeply about unbridled AI.

Northwest · M

@Ontheroad

want AI to move forward with all the possibilities it gives us

This is the source of the misunderstanding.

AI is not a single thing.

The Agent does stuff. For instance, figure out what an email to me should have in it.

The Model makes the model generate an email to me

The Memory determines how long the path the Agent took and what the Model did live.

The Workflow determines how many Agents and perhaps Models are involved and may also control Memory.

The RAG may use an base Agent/Model and figure out how to apply to me, and starts (based on memory) keeping track of my Active Nodes

We want some of these things, for instance the Agents and the Models, but the risk of Workflow, RAG and Active Nodes going off the rails is where we stop and listen.

Ontheroad · M

@Northwest what was at play is that AI is many things... something you noted. What it is to you and what it is to me or Jane or Jim down the street are different things based on what we use (or don't use) AI to accomplsh.

We are still in its infancy stage and careful is for sure what AI calls for... on all fronts and in all its many forms and uses.

ArishMell · 70-79, M

I wonder what the Pope, who is no ignoramus about technical matters, actually meant by his metaphor; but the world certainly need heed his warning. Increasingly power is being taken, vary dangerously and unaccountably, by the small number of extremely wealthy men who own the AI and major social-media companies.

View 1 more replies »

ArishMell · 70-79, M

@Northwest If it ever reaches that level, it won't be the electronic servers but their human owners and programmers who would be to blame

Northwest · M

@ArishMell Looking for blame would be too late. Hence the need to end it before it starts.

ArishMell · 70-79, M

@Northwest I agree wholeheartedly. It needs the foresight and courage at governmental and technical level to do so.

Add a comment...