Cracking the code - why people do matter more than AI! - diginomica

The AI sector’s attention over the past week has centered on the fallout from Anthropic CEO Dario Amodei’s dramatic declaration that he and his peers need to take their metaphorical foot off the tech accelerator pedal and pause for some ethical and safety thought.

But while Amodei - along with OpenAI’s Sam Altman and NVIDIA’s Jensen Huang - were at Salesforce’s Dreamforce jamboree to field further questions on this, Microsoft quietly slipped something rather pertinent to debate out the door.

It’s a humanist Code of Conduct for AI, a 37 page draft opened up for public comment for six weeks, that addresses many of the ethical and safety concerns that have been advanced around the speed at which the sector is racing hellbent in pursuit of the goal of so-called superintelligence, seemingly at any cost. It’s come out of Microsoft.ai, a research lab with a mission statement on its website to the effect that:

We are building a practical superintelligence, rooted in the challenges of real people and businesses. This is the world’s most advanced AI, designed to address our biggest problems and always remain under human control.

According to the firm:

The speed of AI development is accelerating. Systems get dramatically more capable every few months. AI adoption and usage continues to increase. The recent safety incidents of large-scale, highly-co-ordinated, and persistent hacking campaigns of AI agents prove that there’s no time to waste. The stakes are high and only getting higher. We believe it is more important than ever to create safe and reliable AI in service of people, and to be transparent about how we go about it.

As part of that, the Code of Conduct is intended to provide a North Star against which to evaluate and train AI, starting from a crucial first principle - people matter more than AI.

Say that again - People. Matter. More. Than. AI!

In its documentation, micosroft.ai expands on this self-evident truism, declaring:

AI should be a tool, not a person, and should never resist being switched off. It should make people feel healthier, happier, and more productive. It should expand human potential and boost living standards, helping people and organizations achieve more than they ever thought possible.

What does the Code say?

The Code sets out to lay down guidelines to ensure that AI models never resist human intervention, correction, or shutdown. Nor would they take it upon themselves to widen their remits to pursue goals that have not been given to them by their human operators.

And there are what Microsoft.ai terms ‘Absolute Constraints’, which take in obvious no-go areas such as mass harm, child safety, and harmful manipulation.

There are ten declarations to be found therein.

(1) People matter more than AI.:

Technology’s purpose is to advance human civilization and to accelerate human flourishing.

(2) The idea of model welfare is wrong. AIs should not have rights like human beings do. This is a welcome stance against insidious anthropomorphism. The Code argues:

[AI] is not conscious and should not be designed to imitate consciousness. It should be engineered to avoid representing as though it has feelings, subjective preferences, or intrinsic motivation. We reject the pursuit of legal personhood, or the idea that models might deserve welfare, or be entitled to rights.

(3) A model should not be able to meaningfully violate this Code of Conduct.

(4) Given a choice between finishing a job or stepping beyond the terms of the Code, it should fail the job.

(5) Slow down! Rushing into building superintelligence that can slip its own leash is not the goal:

[Humanist AI] rejects the race to produce an all-purpose superintelligence that could evade these safeguards. We are building something fundamentally useful and safe even if that means compromising on ultimate generality, autonomy, or capability.

(6) A model must be interruptible, correctable, shut-down-able. If it isn’t, don’t ship it!

(7) No neuralese. If humans can’t understand it, humans can’t oversee it.

(8) AI should make humans sharper, not dependent:

Models should discourage patterns of interaction that cause excessive reliance or emotional dependence.

(9) Pluralism, yes. Moral relativism, no:

Pluralism does not mean neutrality toward harm or that anything goes. Humanism is rooted in broad, inclusive human values such as human dignity, safety, autonomy, and fundamental human rights.

(10) Humans must be clear about what AI must never do as about what it will do:

Defining what they must not do is as important as our excitement and optimism about the tremendous benefits they’ll bring.

The time is here

CEO Mustafa Suleyman argues that the Code is needed now, declaring:

The time to act has clearly arrived.

To address this need, Suleyman last November set out what he called ‘Humanist Superintelligence’ The Code of Conduct is built out from that. By his own admission, much of the Code is negative and about what AI shouldn’t be able to do. But there are positive aspects, he insists:

Our AI should leave people more capable. It shouldn’t compromise on human autonomy or judgement but support them…We want to create incredible AI, but not superintelligence at any cost. That’s an anti-goal for humanist AI.

He concludes:

The time to have answers is running out. I hope this is, for our part, a start.

My take

It is.

The release of the Code comes shortly after Microsoft CEO Satya Nadella offered his view in the wake of the Anthropic declaration:

Any pursuit of superintelligence has to be grounded in the core principle that if the AI we build is not helping humanity and under human control, it's not worth pursuing.

We also need to accelerate and spread the benefits of AI, such that they are diffused broadly across countries, communities, and companies. This requires a frontier ecosystem in which both closed and open-source models can thrive.

And for firms, it’s imperative that they retain full control over their unique and tacit knowledge. Every organization should be able to build its own continuous learning loop/hill climbing machine, without becoming dependent on any one model provider, and have the ability to embed its own knowledge into models and weights they control.

With that said, he welcomed what he called the “deliberate pacing needed to get alignment right”, as well as embracing Amodei’s idea of third party "embedded evaluators" as well as broader efforts to “develop the mechanisms to make this more than just talk”:

The key is that this cannot be controlled by a handful of entities, but must have broad representation across the ecosystem, countries, and fields, including academia.

All told, a useful contribution from Microsoft to the wider febrile debate that’s dominated over the past few days.