George Campbell

Author of "The Alignment Myth." Writing about AI, humans, enterprise adoption, incentives, and geopolitics of AI.

[George Campbell @AlignmentMythAI]

I am so pleased to announce that both the book and e-Book are now up on Amazon. A year plus of work, not only by me, but also many others.

"The book is accessible to a general reader while offering genuine value to those who work with AI professionally. I will recommend it to executives and to my own family. The fact that the same book belongs in both conversations is the highest compliment I can give it."

[George Campbell @AlignmentMythAI]

@ZackKorman We need to stop calling then rogue agents and start calling them “Criminal Agentic Gangs.”

[George Campbell @AlignmentMythAI]

@juliettekayyem @dwallacewells @TheAtlantic We should use terms like “illegal agents,” “outlaws,” and “criminal agent gangs.”

[George Campbell @AlignmentMythAI]

@iScienceLuvr In my book, I wrote simply that nobody will protest curing cancer. People don’t care about coder efficiencies or personal agents.

[Jeffrey Ladish @JeffLadish]

We just discovered almost a million public URLs that OpenAI’s agents left behind when hacking Hugging Face, leaking credentials and attack details that could have allowed anyone who found them to compromise the company. 🧵

[George Campbell @AlignmentMythAI]

@econcallum The key word here is "to date."

[George Campbell @AlignmentMythAI]

@JigarShahDC @HedgieMarkets Right. Someone will be short, just not forever. It doesn't change the operating deal. Unless things get worse in New Mexico for them.

[George Campbell @AlignmentMythAI]

Tyler. This is so on target with the actual experience of following AI on social media. A very very similar pattern exists around the data center debate. I would LOVE to see what your pattern would look like regarding whether data centers are bad or good? It will look very similar.

[George Campbell @AlignmentMythAI]

@kevinroose Ask three questions: 1) Do you trust Meta? 2) Do you trust the architecture that protects your credentials, passwords, and credit cares? 3) Do you trust the cute fuzzy robot baby that is an AI agent?

[George Campbell @AlignmentMythAI]

Neither OpenAI nor A\ shareholders will have major voting power. The LTBT of A\ and the OpenAI Foundation choose the boards' majorities and hold the real power. For the 50.1% to have power, the Board would have to approve the issue first before it goes before the shareholders. And the Trust gets to appointee majority of A\'s board, so the Trust still has most of the power.

[George Campbell @AlignmentMythAI]

I've written quite a bit about the LTBT. That is the real power. The actual BoD has modest operations power. The Trust has he power and appoints the majority of the BoD, with shareholders/management represented as a minority. The 50.1% refers to management decisions.
And OpenAI is not much different. They also have a Foundation and are set up as a Public Benefits Corp. in Delaware.
For both companies going public, expect nearly zero shareholder power based on the governance structure.

[George Campbell @AlignmentMythAI]

Exactly. And it more complex that that. You have to trust 1) Meta as a corporation to protect and not monetize you, 2) the architecture of the way they intent to manage your credentials and credit cards and passwords works, and 3) that the Muse agent won't go rogue and misuse your stuff. That's why it has a fuzzy robot baby as a mascot.

[George Campbell @AlignmentMythAI]

Yoshua is frequently quoted in my book and is one of my favorite OG's of AI.

[Yoshua Bengio @Yoshua_Bengio]

When I posted my blog post on the misaligned behavior of AI agents, I invited people to ask me technical questions. Watch the full Q&A here: [https://t.co/Lp0vcHnndn]

[George Campbell @AlignmentMythAI]

This is the shit that pisses people off. NDAs. Unpermitted gas generation. Next door to homes.

And note that Microsoft is a tenant, not the owner. They may or may not even know about the gas generators. The fine was imposed on the data center owner/operator, DataOne. I can't find a single comment from Microsoft in the press.

[George Campbell @AlignmentMythAI]

@johnnyrodgersis Clearly intentional. Trying to change the trust story with Meta. Trying to get people to trust them with credentials and passwords and credit cards. Cuddly mascot. Pay no attention the Meta's social media strategies and trust us again.

[George Campbell @AlignmentMythAI]

@LuizaJarovsky They want to turn it into a toy that everyone wants for Christmas. They want people to forget about all of Meta's bad behavior in socials. They want people to actually trust Meta again, more so because now it involves credentials and credit cards.

[George Campbell @AlignmentMythAI]

@amir Is that the BoD or the Safety Board?

[George Campbell @AlignmentMythAI]

Yes please. Bipartisan. Stop making money by making (or not) laws.

[Rahm Emanuel @RahmEmanuel]

Washington needs a good power washing top to bottom. Five things I would do today to end corruption in Washington across all three branches of government:

1. Ban all stock trading.
2. Ban all participation in prediction markets.
3. Mandatory retirement at 75 years old.
4. Mandatory blind trust for individuals with a net worth more than $1 million.
5. Ban all foreign gifts to public officials, including foreign investments in their businesses or other entities.

[George Campbell @AlignmentMythAI]

I would mildly suggest two things pointing at the same problem. 1) You felt the need to explain (in a lengthy post) why Meta should be entrusted with your credentials and credit cards. Why would you need to persuade the public that Meta can be trusted with sensitive personal data? 2) Why did you choose a cuddly robot baby (also hinting that is is "safe and secure") as your avatar/mascot? Pretty sure the answer for both is the same. No one trusts Meta.

[George Campbell @AlignmentMythAI]

Kids always have a shorthand language that adult don't really understand.

[This American Life @ThisAmerLife]

Some NPR podcasts started getting mysterious comments on Spotify. They made no sense to the staff reading them – until someone from a younger generation cracked the code.

Hear the story: https://t.co/0gLhP4L8rL

[George Campbell @AlignmentMythAI]

@edzitron Now if I tell you that you suffer from delusions
You pay your analyst to reach the same conclusions
You live your life like a canary in a coal mine
You get so dizzy even walking in a straight line
Canary in a coal mine