15 Comments
Dec 11, 2022Liked by Samuel Hammond

> Indeed, while ChatGPT doesn’t yet pass the Turing Test, it easily drafts an essay on all the ways AI reinforces the patriarchy.

I’m particularly excited, and frightened, by the potential for ChatGPT-like methods to create novel political arguments for persuading voters in unexpected ways.

For example, “Why should a Trump voter switch to supporting Warren?” To a human, that seems like an academic exercise or a WaPo article that would fail to convince anyone. Yet an AI program may have deeper insight into how specific types of voters think and the language that resonates with them. E.g., it may craft its persuasion as an attack on Biden that also introduces Trump supporters to left-of-center criticism using Republican-friendly language and values.

Further, AI persuasion needn’t just take the form of a single article. Instead, an AI chatbot could chart a long war of persuasion through months of interactions. It could identify particularly susceptible targets and begin to regularly engage them in contexts like Twitter replies. The AI would adapt based on each person’s responses and slowly influence the target’s thoughts and beliefs towards the AI’s objective. Through experimentation with millions of people, the AI would continuously learn more effective persuasion tactics.

Expand full comment

I'm not as sanguine about lawyers' future as you seem to be. I think the best lawyers will move up the value chain and continue to make $$$ but I think AI will render the commodity lawyer obsolete.

Expand full comment
Dec 11, 2022Liked by Samuel Hammond

> In the medium term, the single best way to capitalize on trends in AI is to start a company.

I think it will be challenging for AI startups to develop a defensible moat. As many analysts [1] and VCs [2] have pointed out, incumbents like OpenAI and Google benefit from their massive infrastructure investments in AI hardware and data collection/curation. Therefore many startups are just outsourcing the AI challenges to these platforms. For example, using OpenAI’s GPT API. Competitors could easily duplicate that since none of these startups have any proprietary AI tech edge.

Additionally, the incumbent platforms could choose to offer these applications themselves to capture more value and therefore restrict the API to eliminate competition. There’s an analogy in how numerous startups were built on top of Facebook or Twitter about a decade ago. These platforms eventually limited API access to keep users on their site, thereby capturing more ad revenue. [3] The startups failed.

Even when startups don’t rely on a platform, they still end up using commoditized methods. For example, the popular iOS application Lensa uses the Stable Diffusion model. Since the model is open source, there have been numerous competitors using the same approach. [4] Only well-capitalized firms like OpenAI and Google can make the investment necessary for proprietary AI tech that offers a defensible moat.

[1] https://stratechery.com/2022/the-ai-unbundling/

[2] https://youtu.be/4qOEg4LbdTU?t=4205

[3] https://techcrunch.com/2012/08/16/twitter-api-client-apps/

[4] https://techbriefly.com/2022/12/07/lensa-ai-free-alternatives/

Expand full comment

"As in other cases of creative-destruction, commoditized intelligence will radically change the economy’s cost structure, producing the institutional churn needed to bust monopolies, dissipate economic rents, expand wealth and opportunity, and right-size institutions."

I hate to be this guy but this seems extremely naive don't you think.

Expand full comment

One key point for AI is being used for democracy as for liquid democracy and public systems co-creation. However, AI can also be used for digital dictatorship or "digitatorship" as in China. https://accidentaleuropean.com/tripadvisor-inspiration-for-our-digital-democracy/

Expand full comment

The thing with banking is that the AI is not the main disruption or alternative - it is crypto. Because AI is not always about decentralization but crypto in principle is.

Expand full comment

Great piece but 1 problem w/ ideas here: output that is convincing but wrong is still an unsolved problem with at least publicly available models.

Would make some educational use cases suspect.

Expand full comment

Why isn’t this advice about investing in land, etc not refuted by efficient market hypothesis? AI is not a secret and so the market should have figured this stuff out already? The part about just going into mutual funds for that reasons contradicts the rest of the post.

Expand full comment

How would one create a long position on AI with companies that already exist in the US stock market?

Expand full comment

What a creepy teddy bear! But probably a really soft and cushy lining is the only way for an AI to survive a one-year-old. However much the kid throws Teddy AI on the floor and bangs it into the wall and stomps on it, its inner AI will survive. And what precedence does that set for the kid's social learning?

Expand full comment