AI Risks & Rewards — Santiago Bilinkis

Could AI’s ability to make us fall in love with be our downfall? Will AI be like cars, machines that encourage us to be sedentary, or will we use it like a cognitive bicycle — extending our intellectual range while still exercising our minds? 

These are some of the questions raised by this week’s guest Santiago Bilinkis. Santiago is a serial entrepreneur who’s written several books about the interaction between humanity and technology. Artificial, his latest book, has just been released in Spanish.

It’s startling to reflect on how human intelligence has shaped the Earth. AI’s effects may be much greater.

Links

Outline

(00:00) Intro

(2:31) Start of conversation — a decade of optimism and pessimism

(4:45) The coming AI tidal wave

(7:45) The right reaction to the AI rollercoaster: we should be excited and afraid

(9:45) Nuclear equilibrium was chosen, but the developer of the next superweapon could prevent others from developing it

(12:35) OpenAI has created a kind of equilibrium by putting AI in many hands

(15:45) The prosaic dangers of AI

(17:05) Hacking the human love system: AI’s greatest threat?

(19:45) Humans falling in love may not only be possible but inevitable

(21:15) The physical manifestations of AI have a strong influence over our view of it

(23:00) AI bodyguards to protect us against AI attacks

(23:55) Awareness of our human biases may save use

(25:00) Our first interactions with sentient AI will be critical

(26:10) A sentient AI may pretend to not be sentient

(27:25) Perhaps we should be polite to ChatGPT (I, for one, welcome our robot overlords)

(29:00) Does AGI have to be conscious?

(32:30) Perhaps sentience in AI can save us? It may make it reasonable

(34:40) An AGI may have a meaningful link to us in virtue of humanity being its progenitor

(37:30) ChatGPT is like a smart employee but with no intrinsic motivation

(42:20) Will more data and more compute continue to pay dividends?

(47:40) Imitating nature may not necessarily be the best way of building a mind

(49:55) Is my job safe? How will AI change the landscape of work?

(52:00) Authorship and authenticity: how to do things meaningfully, without being the best

(54:50) Imperfection can make things more perfect (but machines might learn this)

(57:00) Bernard Suits’ definition of a game: meaning can be related to the means, not ends.

(58:30) The Cognitive Bicycle: will AI make us cognitively sedentary or will it be a new way of exercising our intellect and extending its range?

(1:01:24) Cognitive prosthetics have displaced some intellectual abilities but nurtured others

(1:06:00) Without our cognitive prosthetics, we’re pretty dumb

(1:12:33) Will AI be a leveller in education?

(1:15:00) The business model of exploiting human weaknesses is powerful. This must not happen with AI

(1:24:25) Using AI to backup the minds of people

Gömböc, a shape at the limits of possibility — Gábor Domokos

The Gömböc is a peculiar shape. It’s too balanced and poised to appear natural but too clunky to suggest it is the fruit of human design. It seems alien or supernatural. This is appropriate, for the Gömböc is a shape that many thought impossible.

View the interactive version.

The Gömböc is the first known homogenous shape with the properties of having just one stable and one unstable balance point. Its weight is evenly distributed yet, however you put it, it will tumble and roll until resting always in the same position. Unless you pose it most delicately on its unstable point.

For convex, homogenous shapes the problem of finding balance points collapses to that of finding local maximal and minimal points from the centre of mass to the surface. For each minimum, there is a stable balance point, for each maximum an unstable one, and for those shapes that have a maximum and a minimum at the same point (appropriately called saddle points, for they are at the top and bottom of “hills”, like a saddle) — it is stable along one direction and unstable along others.

For many years, many mathematicians believed such a shape to be impossible, fruitlessly trying to produce a theorem to prove their intuition. Their hypothesis was informed by two pieces of evidence:

  • For two-dimensional shapes, it can be shown that 4 balance points are the minimal possible. This is proved by the four-vertex theorem: any closed curve other than the circle must have four vertices — four points that are either local maxima or minima.
  • No shape had been seen in nature.

Evidence can be misleading.

This week Gábor Domokos relates his decade-long quest to find the Gömböc. A tale of mathematical intuition, persistence, and a dose of luck.

But the story of this shape is far more than a mathematical curio.

Its discovery developed ways of thinking and led to a research program that has revolutionized our understanding of how things fall apart. Indeed, the Gömböc and the cube can be understood as two bookends of the evolutionary process by which material disintegrates.

Whether rocks on Mars, pebbles in Pisa, or asteroids flying through space, the Gömböc and the cube delimit the forms that things can take. At the one extreme, the cube sets an upper limit on (or, rather, just above) the number of balance points natural objects take.

Take a large boulder, or block of ice, and fracture it with a pickaxe, the resulting pieces will tend to have a similar number of faces as the cube. Just as randomly drawing lines on paper will produce, on average, shapes with four sides — see the image below, or animation here. As rocks weather, the areas of the highest curvature (the pointiest bits) weather fastest. This will tend to reduce the number of balance points as faces disappear and points of maximal and minimal distance elide.

The Markovian march towars the Gömböc from Pebbles, shapes, and equilibria Gabor Domokos, András Árpád Sipos, Tímea Szabó & Péter Várkonyi

These straightforward techniques for categorizing forms in terms of balance points, average number of faces (and also of vertices, and faces per vertex) have proved incredibly productive. The Platonic pantheon of shapes (tetrahedra, cubes …) and the few others we learn of at school (cones, cylinders … ) are but a few citizens of a world that comes into view. A jostling populace that ineluctably marches towards the Gömböc.

The extraordinary consequence of this is that holding a pebble in your hand you can feel its age, not just in its sea-worn smoothness, but in its geometrical simplicity. In the loss of its stable points.

Things fall apart

View animated full screen version

Reading

Moments

(00:00) Intro

(2:40) Start of conversation — what is a Gomboc?

(4:30) The Gomboc is the “ultimate shape” it has only two balance points

(5:30) The four vertex theorem: why a 2D shape must have 4 balance points

(6:30) (almost) nobody thought a Gomboc existed

(8:30) Vladimir Ilych Arnold’s conjecture

(9:00) Hamburg 1995, the beginning of a quest

(10:30) “Mathematics is a part of physics where experiments are cheap”

(11:50) A hungry scholar sits next to a mathematical superstar

(13:00) Ten years of searching

(15:00) Domokos and Varkonyi’s gift for Arnold

(15:30) Arnold’s response: “good, but now do something serious”

(16:50) We cannot easily speak about shapes.

(18:00) A system for naming shapes

(21:00) “The evolution of shapes is imprinted in these numbers”

(21:50) Pebbles evolve towards the Gomboc, but never get there

(24:50) How to find the balance points of shapes by hand

(30:00) Physical intuition and empirical exploration can inform mathematics

(30:30) A beach holiday (and a marital bifurcation point)

(34:00) “No this was not fun, it was a markov process”

(36:40) Working with NASA to understand the age of martian pebbles

(38:20) An asteroid, or a spaceship?

(43:00) The mechanisms of abrasion

(45:50) The isoperimetric ratio — does not evolve monotonically …

(47:50) … But the drift to less balance points is monotonic

(49:00) The process of abrasion is a process of simplifying

(50:00) We can name the shape of Oumuamua because it is so simple

(51:00) Relationship between Gomboc and (one way of thinking about) entropy

(55:00) Abrasion and the heat equation — curvature is “like” heat and gets smoothed out

(58:00) The soap bar model — why pointy bits become smooth

(1:00:00) Richard Hamilton, the Poincaré conjecture and pebbles

(1:04:00) The connection between the Ricci flow and pebble evolution

(1:09:00) Turning the lights on in a darkened labyrinth

(1:12:00) The importance of geometric objects in physics (string theory)

(1:13:30) Another way of naming natural shapes: the average number of faces and vertices

(1:15:00) “Earth is made of cubes” — it turns out Plato was right

(1:16:30) Could Plato’s claim have been empirically inspired?

(1:17:50) “Everything happens between 20 and 6”

(1:18:30) The Cube and the Gomboc are the bookends of natural shapes

(1:19:30) The Obelisk in 2001 — an unnatural, but almost natural shape

(1:22:00) Poincaré on dreaming: genius taps the subconscious

What is philosophy? — Simon Critchley

From what human need does philosophy emerge? And where can it lead us?

Simon Critchley is Hans Jonas professor of Philosophy at the New School in New York, and a scholar of Heidegger, Pessoa, Football (Liverpool FC), humour. He crosses over between analytic and continental traditions and freely draws on quotes from Hume and British pop bands.

Simon argues that philosophy begins in disappointment, not wonder. But it does not end there, its goals can be wisdom, knowledge, enlightenment, and freedom. Its concerns can be yet more varied: it can work as a tool for developing scientific theories, for exposing ideology, or for tracing the underpinnings of language and experience. Anywhere where other fields fear to tread, that’s where philosophers step in.

Since recording with Simon, one question has kept turning over in my head: how important is context to the understanding of philosophical ideas? An issue we discussed towards the end of the interview and did not quite have sufficient time to loop around as many times as I would have liked.

On the one hand, I believe in such things as facts and truth. I believe that science and some philosophical reasoning can deliver us these — although I recognize that belief cannot be justified beyond doubt. It is an act of faith. However, certain types of things, for example, facts concerning the laws of physics, I take to be impervious to our attitudes and human concerns. Why we believe them, and why we even look for them, are valid sociological and phenomenological questions. Yet the evidence for these facts can be produced usefully without any reference to historical context and the facts themselves do not depend on human factors.

In this way, to the question of whether the sun existed before humans — the subject of a heated discussion between Maurice Merleau-Ponty, Georges Bataille and A J Ayer in a Parisian bar — my answer is yes.

And yet.

There are questions that philosophy asks about issues that have no basis in fact. Impossible objects.

I recently asked ChatGPT to produce some predictions about the future for a project I am working on (What Year is Now? A Hallucinatory Horological Mapping). There are some obvious errors with the below — the tercentenary of the first interdimensional travel falling in 2307 would imply we first made that leap back in 2007. I suppose that’s possible if the dimension in question is time.

More interesting is what is predicted for 2306:

2306: Interdimensional exploration reveals the nature of existence itself

ChatGPT

I do not think that will happen. I don’t think it can. Determining the nature of existence is a question we cannot solve, even though there is value in turning it over, the value is not to get to an answer. It’s a means justify the ends (and btw there are no ends) sort of question.

For questions like this, and more particularly questions about the meaning of existence, context is important. Consider also questions in ethics and meta-ethics. There are no truths to be had here, and what is right for society — perhaps even the right way of thinking for society — will depend on that society. For example, in times of high uncertainty, where predictions are costly and inaccurate (perhaps this moment?) we might have reasons to consider virtue ethics over consequentialist frameworks.

Does it matter that I write this on Sunday morning, in 2023, in a stony city, that the sun — prehistoric or not — is hidden by several layers of cloud? It’s not for me to say.

Some links

Moments

(00:00) Intro
(3:00) Beginning of conversation: disappointment as the start of the journey
(7:55) Punk & Philosophy
(11:20) Trauma and tabula rasa
(12:30) Not making it in a band, becoming a philosopher
(19:30) Wittgenstein as a bridge between analytic and continental philosophy
(21:50) Mill and the origin of the label “continental philosophy”
(24:30) Philosophy has a duty to be part of culture
(28:00) The difficulty with philosophy being an academic tradition
(29:30) The Stone
(32:30) Football as a phenomenon for study that invites people in to philosophy
(35:00) Philosophy as pre-theoretic & Pessoa’s Ultimatum
(39:00) Will analytic philosophy run out for road and be subsumed into science?
(41:00) Two lines of human imagination
(42:00) Should philosophy ever be a single honours subject, or should it always aid other realms of thought?
(43:00) Philosophy as pre-science
(44:30) Phenomenology as reflection on the lived world
(47:00) Alberto Caeiro (Pessoa) and anti-poetry
(48:50) The saying of ordinary things to fascinate angels
(54:00) Impossible objects will keep philosophers busy
(57:00) The task of philosophy as deflationary, as not making progress
(1:00:00) Should philosophy of physics be part of physics?
(1:04:30) Context: What can’t I read Descartes like I’m talking to your right now?
(1:06:00) Is context colour or is it inseparable from ideas?
(1:15:30) Rorty: Continental philosophy as proper names vs problems in analytic philsophy
(1:19:20) Trying to walk the line between two traditions of philosophy
(1:20:00) Obscurantism vs scientism
(1:23:00) Permission to think on their own, to expose ideology
(1:26:00) The internet has been good for philosophy
(1:26:30) Audio as a new platform or agora for philosophy

ChatGPT as a Glider — James Intriligator

Large language models, such as ChatGPT are poised to change the way we develop, research, and perhaps even think (see The Offshoring of Thought and Memory). But how do we best understand LLMs to get the most from our prompting?

Thinking of LLMs as deep neural networks, while correct, is not very useful in practical terms. It doesn’t help us interact with them, rather as thinking of human behavior as nothing more than the result of neurons firing won’t make you many friends. However, thinking of LLMs as search engines is also faulty — they are notoriously unreliable for facts.

Some other models have been proposed:

  • LLMs are “stochastic parrots” as Bender, Gebru, McMillan-Major, and (Sh)Mitchell argue
  • ChatGPT is “A fuzzy JPEG of the web” according to Ted Chiang

These both capture something of how they work, but they do not provide any direction on how to create prompts.

Our guest this week is James Intriligator. James trained as a cognitive neuroscientist at Harvard, but then gravitated towards design and is currently Professor of the Practice in Human Factors Engineering and Director of Strategic Innovation at Tufts University. 

James proposes viewing ChatGPT not as a search engine, parrot, or JPEG, but as a “glider” that journeys through knowledge. By guiding it through diverse domains, it learns your interests and customizes better answers. Dimensional prompts activate specific areas like medicine or economics. 

I believe we’ll need to have various mental models to understand how best to interact with LLMs. This is one for the toolbox.

Links:

Generative Art using GPT4 #2: 3D fractals

Click for full size, interactive version

There’s a serendipitous quality to experimenting with LLMs.

I was trying to make an interactive model of a Mandelbulb with ChatGPT. Although it didn’t work as intended, it produced something funky. Indeed, for making art their fallibility might be a feature and not a bug.

Here’s a tip for making HTML5 from code generated by LLMs:

  • It will get stuff wrong — frequently things won’t render at all
  • Paste the results into a text file, save as .html, open in Chrome, open the Developer Tools Panel in Chrome
  • If you have any errors they will be listed (see below) paste them into Chat GPT or your LLM of choice. Let it figure out what went wrong.

Generative Art using GPT4 #1: fractals

Mandelbrot Fractal Zoom

To produce something for many people was once an expensive endeavor, capital was required to open a factory or obtain a printing press.

If you’re reading this, you probably own a factory and a printing press. It’s the object attached to this screen. But you might need workers to help you turn ideas into products. This is where LLMs can help.

I’ve long been fascinated with fractals, these entities so rich in detail that they seem to live between dimensions as if overspilling the plane but not filling the volume. But I’d never bother coding them, the idea was enough, the work was too much.

With ChatGPT4 I created the above HTML5 fractal in a few steps, starting with the prompt to use the Multiverses palette:

Draw a Mandelbrot fractal using the palette #A66021 #07091C #19837E #11C55A #FFFD34 #B9FB4F #732401

Using the Code Interpreter this, pleasingly, produces an image in the ChatGPT window using Python. A few more prompts created one in HTML5 that zooms on click with overlay text that disappears.

What I’m feeling right now is a fractal sense of accomplishment — I’d be more proud if I’d coded this from scratch — to have been an artisan rather than a foreman. I’ve overspilled the plane, but not filled the volume.

Phylogeny & The Canterbury Tales — Peter Robinson

The physical solidity of books encourages notions of “the text” or “the canonical edition”. The challenges to this view from post-modernist thought are well known. But there are other ways in which this model of a static text may fail. Our guest this week is Peter Robinson (my dad!) who takes us through his work … Read more

MV#11 — AI, risk, fairness & responsibility — John Zerilli

AI is already changing the world. It’s tempting to assume that AI will be so transformative that we’ll inevitably fail to harness it correctly, succumbing to its Promethean flames. While caution is due, it’s instructive to note that in many respects AI does not create entirely new challenges but rather exacerbates or uncovers existing ones. … Read more