16 Comments

This military money might bring advancements which benefit the wider population as well. Maybe...

Expand full comment

I'd say we go back to the proposed idea of having nuclear launch codes in someone's chest cavity. That's a good way to keep a human always "in the loop" , but in all seriousness it seems kinda ironic that possibly one of the safest options would be to "analoguize" (in case it isn't already I'm.not knowledgeable enough) most of the chain of command and capacities of a nuclear launch. But then again misinformation inside the organization could provide a human failure in the system ... 🤷‍♂️

Expand full comment

Human error is the cause of most unintended outcomes and accidents.

Expand full comment

Would you consider bias in algorithms a form of human error? Would you consider all white male venture capital rooms Human error? I recently just read another Asian who is fired from a venture capital firm with the usual complaints. If the same things keep happening, and we project all of this into even more powerful artificial intelligence? I'm curious how human error will compound?

Expand full comment

Yes, definitely.

Funny, I was going to throw in a question to ask if you're going to address the inherent gender bias of AI tools. 😁

Expand full comment

It's gender bias in Venture Capital that compounds to everything as well. It also significantly hasn't changed much in the last decade: (I know we think a like)

https://techcrunch.com/2023/05/01/ann-lai-says-she-was-fired-from-bullpen-capital-after-helping-deliver-a-145m-fund/

Expand full comment

I was thinking of gender bias in IT itself, which includes the new beaut man made AI tools.

Don't forget that even facial recognition was initially built with a Caucasian bias, yet no one noticed or thought of it until facial recognition had trouble reading non compliant skin colours!

Gender bias in venture capital didn't create that problem, nor would more or differently balanced capital investment have prevented the problem.

The blinkered thinking is real.

Expand full comment

Maybe A.I. can help corruption from seeping into our political, national defense and leadership arenas. Beginning to think that the game is fixed on multiple levels to profit the few. Keeping "humans in the loop" means having people in charge who are full of bias, prone to errors, able to make bad decisions and have policies that benefits a privilege few, while being to the detriment to others.

The latest regional bank failures is such an example of failed policy and regulation at scale. A.I. should be made into a custodian "intelligence" to oversee humanity. So that we don't destroy ourselves or exploit each other to the degree that we are on track to do so.

Expand full comment

I get your point Michael, but I have issues with the concept of a "Nanny" AI keeping humanity safe from ourselves when the AI itself is coded and built by biased humans themselves. And also it all goes back to alignment when you give a utility function based on keeping humans safe how do we avoid a prison or any other "efficient" ways of keeping us safe without extinction 😂

It can get pretty dicey pretty fast. Idk I still think a possible solution could arise in the foreseeable future in the meantime it's fun to watch it all unfold.

Expand full comment

And interesting those that oppose antitrust regulation often make the national security argument: Facebook was the main donor to a group that fought antitrust reforms in 2020 and 2021

https://www.cnbc.com/2023/05/01/facebook-primary-donor-group-antitrust-fight.html

Expand full comment

Disappointed in this one. You object to Silicon Valley’s AI hype, yet your own Terminator headline and talk of ASI is equally breathless. You are stoking fears just as you say SV is stoking AGI to sell products.

No, AI-enabled killer drones are not imminent. Yes, there are people in the US military who actually care about responsible AI. Yes, there are guardrails and rules in place in the US military to ensure that AI is used in safe, lawful, and ethical ways. More so in fact than in some commercial companies.

You do your readers a disservice by talking about military AI in ways that are not matched by the facts. If people really understood the state of AI in the US military today, they would likely be underwhelmed. Not running for the hills because Terminator might be just around the corner.

Expand full comment

Regardless of hype or not, military AI is such a sensitive issue that we need to discuss it and draw the line somewhere before it is too late.

Expand full comment

Perhaps you can give this lecture to Stephen Hawking or Geoffrey Hinton Jack: I don't take well to this kind of personal attack without you backing up what you're saying with facts or links or something. Do you think the proposed Bill will change anything? Did you read what Hinton said as he's leaving Google?

Expand full comment

It’s not a personal attack. I didn’t mean it to sound like one. It’s based on five years of working on AI programs for the DoD. Practical, daily, hands-on experience. Not theory. I know what exists and what doesn’t exist. And I know many people are working hard to get it right.

For all the talk about killer drones, I worry much more about the near-term consequences of generative AI for misinformation and disinformation. And general societal malaise. That problem is upon us today. Right now. The people developing these capabilities now see the potential harm these systems can cause. But that’s not been the focus for the past five years.

Expand full comment

And by the way Jack, such an EPIC lack of foresight, the moment the DoD brought in Google to Maven, it made the CCP think about how its own BigTech and facial recognition companies should work directly with the PLA. A pioneer in AI regulation, at the DoD? Maybe singularly responsible for some of this corporate - national security arms race. That's on you Jack.

No wonder you blew up at me. You want to talk adversarial, you don't even seem to realize the consequences of Maven in the world today.

https://thegradientpub.substack.com/p/lt-general-jack-shanahan-ai-in-the#details

Expand full comment