Welcome to another issue of CTPI Insights, your premium newsletter on technology design and policy.

Today, we'll cover:

Biden's AI Executive Order;
Ben Thompson's justified fury; and
Our take on what's going on… from 5 months ago.

Hello again, friends. We have some exciting news. As CTPI evolves as a company, we're making a few changes in our services.

The first important announcement is that we're pausing the premium CTPI Membership for a while. We thank so much each of you who supported us through these first months. We'd love to name each of you, but you know… we value your privacy. Starting next week, CTPI Insights will be free for all. We'll continue to explore ways of delivering more value for you, and we'll let you know once our new version of a premium service is ready for purchase. Until then, we'll continue delivering great biweekly content to all of you, and we encourage you to share our content with as many people as you can.

We're also expanding our CTPI team. So you'll probably start seeing other amazing people's work around here, too. More importantly, we'll double down on our high-quality research and exciting new projects to apply the knowledge we are developing here with all of you.

Let us know if you have any specific questions, and let's change the world together.

Now, let's get to it.

M. and J.

The US AI Executive Order

From TechCrunch:

"U.S. President Joe Biden has issued an executive order (EO) that seeks to establish “new standards” for AI safety and security, including requirements for companies developing foundation AI models to notify the federal government and share results of all safety tests before they’re deployed to the public.”

This piece from the Financial Times gives more context about the kind of action the Executive Order entails:

“The White House will use the Defense Production Act, a cold war law used at the peak of the Covid-19 pandemic, to compel businesses developing AI models posing serious risks to national security, economic security, or public health to notify the government when training these systems and to share their safety test results”.

The AI Executive Order, however, takes a technology-specific approach and threatens the health of technology innovation for the future.

Ben Thompson wrote a killer piece about the AI EO explaining exactly that. It's worth reading the whole thing, but we'll highlight two crucial points.

First, he recalls the big AI companies were the ones calling for regulation in the first place. From Stratechery:

“The point is this: if you accept the premise that regulation locks in incumbents, then it sure is notable that the early AI winners seem the most invested in generating alarm in Washington, D.C. about AI. This despite the fact that their concern is apparently not sufficiently high to, you know, stop their work. No, they are the responsible ones, the ones who care enough to call for regulation; all the better if concerns about imagined harms kneecap inevitable competitors”.

And finally, he brings the point home, analyzing how Biden's approach is the exact anti-pattern of technology innovation. From Stratechery:

“What is so disappointing to me is how utterly opposed this executive order is to how innovation actually happens:

The Biden administration is not embracing uncertainty: it is operating from an assumption that AI is dangerous, despite the fact that many of the listed harms, like learning how to build a bomb or synthesize dangerous chemicals or conduct cyber attacks, are already trivially accomplished on today’s Internet. What is completely lacking is anything other than the briefest of hand waves at AI’s potential upside. The government is Bill Gates, imagining what might be possible, when it ought to be Steve Jobs, humble enough to know it cannot predict the future.

The Biden administration is operating with a fundamental lack of trust in the capability of humans to invent new things, not just technologies, but also use cases, many of which will create new jobs. It can envision how the spreadsheet might imperil bookkeepers, but it can’t imagine how that same tool might unlock entire new industries.

The Biden administration is arrogantly insisting that it ought have a role in dictating the outcomes of an innovation that few if any of its members understand, and almost certainly could not invent. There is, to be sure, a role for oversight and regulation, but that is a blunt instrument best applied after the invention, like an editor”.

Honestly, we're not surprised. We literally wrote about this almost five months ago when we analyzed the background story of OpenAI and the industry's push for regulation. We do apologize for the massive self-quote, but we think we really hit the nail on the head back then, and re-reading it in light of what we know today is quite revealing. From AI Regulation, backstage squabbles, and fear-mongering executives:

“After the massive surge in popularity of ChatGPT, and as people kept arguing AI would be either the redemption or damnation of humanity, regulators joined the conversation. In the US, congressional hearings took place, and Sam Altman went on a world tour to talk to regulators.

The discussion around regulation is of the utmost relevance since it could set the tone for technology development for years to come. And although the regulator's initial plan was to go with a technology-neutral, risk-based set of rules, fear-mongering and other self-interested narratives have successfully changed the negotiations. (…)

This is not by chance. The AIpocalypse narrative is useful for those like Elon Musk, who was left behind in the AI development process. And they are not alone. The more structured push to regulate AI also benefits OpenAI and Microsoft themselves. It's the typical “kicking the ladder” strategy. More regulation elevates entry barriers in a market and tends to shield big-pocket, consolidated players, like the combo OpenAI-Microsoft. That is, of course, precisely the kind of world OpenAI was created to avoid: one in which only a few private companies could hold the future of AI.

Others, like IBM, are going for a delimitation strategy, in which naming several situations where AI could be regulated without much damage for the company can, by contrast, shield applications the tech giant could be targeting in the future. From the Emerging Tech Brew:

“Altman, along with IBM's Christina Montgomery, made suggestions for how Congress should take action, with Altman proposing “licensing and testing requirements” and Montgomery suggesting “precision regulation” of specific use cases.

Microsoft President Brad Smith also threw the company's support behind the creation of a new government agency and licensing system for pre-trained AI models at an event in Washington, DC, last week that laid out Microsoft's blueprint for AI in a white paper”.

Although some are still voicing the risk-based approach solution for regulation, like Salesforce CEO Clara Shih, the statements from the most vocal companies in this debate beg the question: “Who speaks for startups and for a non-monopolistic future?”.

A glimpse into a technology-neutral AI regulation

CTPI is grounded on a firm belief that technology regulation should rarely be focused on the technology itself, but instead on the underlying human activity the new tools are enhancing. We call this a technology-neutral regulation. That means using ChatGPT to cheat in school is bad not because of AI but because cheating is forbidden. In this context of education, from Markkula Center for Applied Ethics’ Susan Kennedy:

“Surprisingly, the panic over ChatGPT doesn't actually seem to be about ChatGPT. It's not all that impressive, nor is it significantly more effective than the ‘old ways’ of cheating. Instead, the panic seems to be fueled by the expectation that students won't be able to resist the temptation to use it and that cheating will become rampant. The release of ChatGPT is forcing educators to confront a much deeper issue that has been taking shape for quite some time; students who are becoming increasingly obsessed with grades, GPAs, and completing a degree, and who are willing to go to great, and sometimes unethical, lengths to achieve these things”.

A technology-neutral approach focuses on the accountability of whoever uses the tool for whatever they are using it for. According to this general logic, using copyrighted material to train AI models shouldn't be a more significant issue than using such tool to replicate a specific artist's style. In the arts, see Maya Ackerman from Markkula Center for Applied Ethics:

"It's important to remember that generative AI has had decades of development within academia. In my experience as a generative AI researcher, I've never seen creative machines in the academic context that focus on imitating the style of a living artist, unless the system was created by that artist. Instead, we have general purpose systems that can create anything from poetry, music, pottery, stories, dance choreography and more. Many academics have even experimented with the idea of AI developing their own unique styles to push the creative abilities of their machines. (…)

The true power of generative AI lies in its ability to elevate human creativity, rather than replace it. In fact, instead of encouraging imitation of other human artists, AI can help creatives find their own unique voice. (…)This is not simply about speed and efficiency, but about opening up a realm of creative possibilities and offering a new type of creative process where inspiration is always abundant.”

Finally, regarding decision-making, as mentioned in a previous issue of CTPI Weekly, individuals and companies should always be held accountable for whatever they say and decide, irrespective of being mediated by artificial intelligence.

Our effort in this discussion must be to strip away our science fiction-based, commercial interests-fulled panic of the technology itself and focus on how to tackle the underlying human actions that effectively afford regulation”.

Curiously enough, it's quite possible that the internal polarization of US politics makes Biden's approach much harder to execute. However, the initial kick in the global AI regulation game was done. And in an entirely wrong direction, as it seems. We'll need to do some serious course correction work if we want to foster open, sustainable, and safe technology innovation.