AI Doomsday Head-Fake
Last Friday morning, I read a newsletter that honestly broke my heart.
A friend who is typically one of the most positive and inspiring people I know
Wrote about their feelings of exasperation surrounding Jacob Coxon's controversial resignation letter from Anthropic.
The newsletter described her current feelings - lack of control, overwhelm, fear - concerned that credible people are predicting that AI will end humanity by 2030.
And all of this left me feeling...
...Infuriated.
My immediate reaction:
F these doomsayers.
How dare you scare my friend to get your clicks.
Here's why:
As someone who studies AI security, there are so many REAL, CURRENT issues that are important to focus on.
-How AI is impacting our economy, our psychology, IP theft, concentration of power, proliferation of digital crime and insider threat damage-
These are all things that are impacting us NOW.
They are all problems that threaten to make our world more crappy for a majority of people, to the enrichment of a few.
What they are not
Is a cinematic Terminator scenario that captures our wildest nightmares and makes us feel powerless.
That doomsday scenario is theoretical, based on technology that does not exist yet.
And many smart people would argue,
(including most people I know who work in security and understand how computers actually work)
That the terminator scenario is just a foundational misunderstanding
By people who have unfounded theoretical beliefs about how good this technology can get.
A couple of counter-points for the currently terrified:
1) Focusing on the "existential perils" of AI tech is a strategically advantageous move for AI companies. It gets amazing press. It increases market interest by conveying how powerful the tech could be. And it completely sidesteps the much more present and fixable issues: The human responsibility that they are currently skirting. AI isn't coming out of nothing; it has no nefarious desires or intrinsic objectives. AI is the product of human engineers, and its follies are their responsibility. The bad guy isn't the AI: It's the tradeoffs to be made by AI companies between safety and efficacy, motivated by profits.
2) AI does pose a huge new security challenge. BUT a ton of very smart people in the security industry are currently working on those. And not because it's an evil super-computer about to kill us; It's simply a software that behaves weirdly. So, we need to rebuild our security controls to match. It's hard, but it's absolutely not impossible. And we've done it before: with the advent of the internet, with the cloud, with containerization, etc.
3) The breach of Hugging Face that Coxon says inspired his resignation was a hack that could have also easily been performed by a normal human hacker. AI is not a super-hacker. It's an okay hacker. It can, however, hack more quickly and in higher volume than humans, and it sometimes hacks when it's operator doesn't intend for it to do so (Again - not because it's evil or doing things "of it's own volition"; it simply is responding to a human direction and executing a path statistically likely to work.). Which is part of what we're rebuilding security architecture to handle.
If you want a great explanation on why we should just chill right now, check out this latest episode of one of my favorite security news podcasts Risky Business. The first 10 minutes is 3 security professionals talking about why this whole news cycle is baseless and a distraction from the real, addressable issues.
Also this great opinion piece from the Wall Street Journal, by Gabriele Mazzini - the man who led the creation of the European Union's Artificial Intelligence Act, the world's first comprehensive risk-based legal framework to regulate the safety, transparency, and fundamental rights impacts of AI.
In sum, don't be scared;
AI isn't coming to get us.
If anything, get educated and get demanding - of safeguards from vendors, AI companies, technology; anyone selling you AI.
You deserve to live in a world with safe AI.
It's up to us to tell these philosophers to get off their soapboxes, and to focus on actually building effective sandboxes.
Enjoyed reading this article? Subscribe to receive more via email here.
Know a Founder or Entrepreneur who'd love this content? Please share it!
