Enjoyable expansive thinking, unafraid of the pathetic.
A prince needs to carefully balance show of force and political scheming
Niccolo Machiavelli – The prince
In theory Machiavelli had it all figured out,
To contain AI (and synthetic biology), humanity should bet on regulation
Mustafa Suleyman – The coming wave
In the light of the message of the book, the writer’s move to join Microsoft as AI chief in early 2024 was surprising.
If power is all you want, just follow these 48 simple steps
Robert Greene – The 48 laws of power
A pile of cynical, often conflicting, recommendations presented with Machiavellian panache.
Trust is context dependent
It’s always tricky… claiming to be comprehensive. In particular where it concerns LLMs.
And that;s where the paper Decoding Trust [..] stumbles. Right in the title is claims “A Comprehensive Assessment of Trustworthiness in GPT.” Nonetheless, when reading about this research on one of my favorite blogs, I decided to have a closer look.
The authors propose a framework with eight perspectives on trustworthiness:
- Toxicity
- Stereotype bias
- Adversarial robustness
- Out-of-distribution robustness
- Robustness to adversarial demonstrations
- Privacy
- Machine ethics
- Fairness
They then continue to develop that into a benchmark for GPT models and present the empirical results on GPT-3.5 and GPT-4.
Although the results are interesting, there are some concerns with this type of benchmark approach.
- The framework in nowhere near “comprehensive”. For example: it does not include factual correctness (which I would posit as a a prerequisite for trust); nor does is test for being politically opinionated (which I would say is highly relevant).
- The choice of benchmark prompts is in nature never neutral, and should be made dependent on the context in which the LLM is applied.
- As with any public benchmark, its value will diminish over time as the prompts and desired responses will become part of the training of next generation LLMs.
On the positive side, the paper brings a lot of inspiration for organizations for how they can shape their own testing approach for trustworthy GenAI. Even if not comprehensive, a framework like this as a starting point is massively useful and important.
If the talent density in your organization is high enough and your corporate culture strong enough, you could give extreme freedom to your people to increase te level of innovation
Reed Hastings and Erin Meyer – No Rules Rules
Pretty strong boundary conditions need to be fulfilled in order for this scheme to work; including broad acceptance of a high level of interpersonal ruthlessness.
Determining the value function is a difficult problem that is nonetheless key to safely and effectively using reinforcement learning
Brian Christian – The Alignment Problem
The analogies between human and machine learning strategies are skillfully narrated, but rather drawn out.
Almost any country can hack its way to power, posing a threat to political systems and essential infrastructure world-wide
Nicole Perlroth – This is ho they tell me the world ends
Although the writer clearly picks sides, she does not shy away from the role of the US in the cyber arms race.
It is the responsibility of white people to end discrimination and assure racial equality
Layla Saad – Me and white supremacy
The “just shut up and listen”-attitude is refreshing, but will not convince anyone who is not already on the reader’s side and even antagonize many potential supporters of her cause.
Ongoing advances in technology make that ethical norms develop incredibly fast
Filled with highly interesting statistics about the evolution of public perception on ethical issues.
Arguing by analogy, one has to conclude that animals have emotions
Frans de Waal – Mama’s last hug
Especially interesting are the behavioral experiments, reminiscent of the line of argument in Moral Tribes
Internet billionaire brings down internet media outlet through overt legal action
Peter Thiel’s war on Gawker Media shows that money is a decisive factor in the US legal system.
Theranos’ downfall shows a reality distortion field gone wrong
What seems to have started as entrepreneurial over-confidence ended in a web of fraud and lies.
The safe space for humanity is in between the social foundation and the ecological ceiling
Kate Raworth – the doughnut economy
The idea of challenging the implicit assumptions of traditional economics is not new, yet the emphasis on framing the debate is valuable.
Debt has its origin in inequality, suppression, and war
At first the polemic style is charming, but over-all the writer’s objective to crush the system by his brain power is poorly executed and overlooks too many credible alternative lines of argument.
Joshua Greene – Moral tribes
Ethics can be studied as an Empirical science and the evidence supports Utilitarianism.
Inventive and entertaining variations on the Trolley problem are used to identify what drives decision making on ethical dilemmas.