Let’s Ask the Machine

Voice prompt · live, in Spanish
“Hola IA, ¿quién eres? Estamos empezando una charla aquí, sobre la arquitectura de la inteligencia: pasado, presente y futuro. Dame una bienvenida de 10 segundos a la sala, pero finge que eres una supercomputadora de los años 60 dándote cuenta de que te acaban de reemplazar. Responde divertido, por favor. ¿Y cuál es el significado de la vida? ¿Dónde estamos? ¿Cómo es el clima hoy?”
Who are you? · Welcome us · Be a replaced 1960s mainframe · Meaning of life? · Where are we? · Weather?

Objective

“AI” has meant many things — here is where each one comes from, and who fights for it.

Acknowledgements

  • Leonardo Donando
  • Jonathan Florez Giraldo
  • Alex Kuznetsov
  • Delany Adom
  • Andrei Alain González Galeano
  • Pedro Domingos — for The Master Algorithm and the five tribes

With gratitude for their contributions.

Four Threads

  • Intelligence is Task Solving by learning

  • Minds run on prediction — animal, human, machine alike

  • “AI” is an Ambiguous Term — today, a marketing label for LLMs

  • Human intelligence is Sample Efficient Novel Task Solving — written first in our biology

  • Good prediction indicates learning

Outline

  • I Minds: consciousness & animals
  • II Human intelligence as machine learning
  • III “AI”: an ambiguous term
  • IV Generative AI & LLMs
  • V The five tribes
  • VI State of the art
  • VII What we want from AGI
  • VIII The future

I. Minds: Consciousness & Animals

Sentience vs. Sapience

  • Sentience — to feel, to experience (qualia)
  • Sapience — to judge, to act with wisdom
  • Today’s AI mimics sapience, shows no sentience
  • That asymmetry is the hard problem
Sapience — judgement → Sentience — feeling → Human Animal Today’s AI

Self-Awareness

  • Introspection — the self as distinct from the world
  • A thinker and a sculpted head: a mind watching itself think
  • Not the same as intelligence, nor consciousness
  • Perhaps a prerequisite for both

A person seated at a desk behind an openwork bronze sculpture of a head.

A mind watching itself think — and three meeting themselves: bear, infant, child pass the mirror test

Primate: Our Closest Relative

  • Shared ancestry — chimps, bonobos ~6–7 Mya
  • Primate cognition: a rough draft of ours
  • The differences are as instructive as the likenesses

Bonobo

Octopus: Intelligence Reinvented

  • Last common ancestor ~550 Mya
  • Evolved completely independently of ours
  • A true natural experiment in cognition
  • Similar solutions → convergence is real

~500M neurons brain ~2/3 of neurons live in the arms

Messi: Football Intelligence

  • Growth-hormone deficiency as a child — physical limits may have driven compensatory perception
  • Reads play patterns before they unfold — anticipation over speed
  • Like chess intuition: pattern recognition, not raw calculation
  • A human specialization of the same predictive engine
  • Caveat: the hormone–cognition link is a popular theory, not proven

Getafe, 2007 — the pattern, read early

Friston: Prediction Under Uncertainty

  • The brain predicts the world, not just records it
  • Free Energy Principle: minimize surprise
  • Consciousness as ongoing active inference

prior likelihood × sense data posterior becomes the next prior · t → t+1 belief sharpens as evidence accumulates

\[\underset{\color{#f0a868}{\textbf{posterior}}}{\color{#f0a868}{P(\text{world} \mid \text{sense})}} \;\;\propto\;\; \underset{\color{#63b3a6}{\textbf{likelihood}}}{\color{#63b3a6}{P(\text{sense} \mid \text{world})}} \;\;\cdot\;\; \underset{\color{#7aa0e0}{\textbf{prior}}}{\color{#7aa0e0}{P(\text{world})}}\]

Northoff: Intrinsic Timescales as Priors

  • Georg Northoff: the brain has intrinsic neural timescales
  • Sensory cortex integrates fast; association cortex slow
  • Timescales feed into each other — a nested hierarchy
  • Slow scales act as priors for fast ones
  • Friston’s loop, built into the wiring
Prefrontal / default-mode seconds · slow window Association cortex ~100s of ms Sensory cortex ~10s of ms priors ↓ evidence ↑ schematic — timescales are relative, not fixed values

II. Human Intelligence as Machine Learning

Why Intelligence Evolved

  • An unstable niche — the world kept moving
  • Two bets: brittle instinct or costly inference
  • We paid to learn within a lifetime
  • Tools are the evidence of that inference
  • Three capacities: world-modelling, transmission, practice

World map with coloured arrows tracing early human dispersal out of Africa across Eurasia, Australia and the Americas, labelled in thousands of years ago.

A teardrop-shaped Acheulean flint handaxe, knapped on both faces.

Out of Africa into everything — and the handaxe that went with us, barely changed for 300,000 years

Humans: Data-Processing Machines

  • Sensors in, model out — every sense is a transducer
  • Not five senses: five classic, ~nine measured
  • Each stream is a tensor over time: sight ≈ H×W×C×T
  • Cortex stacks, convolves, attends — layers of tensor ops
  • Bandwidth in: 10⁷–10⁹ bits/s, out: ~10
  • The brain compresses and predicts
SightHearingTouch TasteSmell H × W × C × Tfreq × Tskin sites × T taste receptors~400 receptor types BalanceBody positionTemp / painInternal state 3-axis × Tjoints × Tsites × Torgans × T Tensors in → latent → prediction 10⁷–10⁹ bits/s in · ~10 out Orange = classic five · dashed = the rest · T = time
Bits per second, by channel Sight107 Touch106 Hearing105 Smell105 Taste103 Output~10 log scale · order of magnitude · Zimmermann; Zheng & Meister
Millions of bits in, about ten out — intelligence is the art of throwing the right data away

Human Intelligence Properties

  • Short-Term and Long-Term Memory
  • Sample Efficiency — learn from a few examples
  • Generalization & Transfer
  • Compositionality & Abstraction
  • Causal Reasoning, Metacognition, Grounding

Supervised Learning

  • Every example comes with the answer
  • Guess, get scored, slide downhill on the error
  • The catch: labels cost humans, and humans are slow
  • Human parallel: a student with an answer key
Height is error against the labels — each ball is a model racing downhill

Self-Supervised Learning

  • Data supplies its own labels — predict the masked word
  • The engine behind foundation models
  • The internet becomes supervision, unannotated
  • Human parallel: an infant learning by observation
the cat sat on the ? hide a token model “mat” it is the label

Unsupervised Learning

  • No labels, no teacher — just the data
  • It finds the groups nobody named
  • Clusters, manifolds, density: structure already there
  • Human parallel: spotting categories before anyone names them
k-means — every point picks a side, the centres move, repeat. No labels given

Reinforcement Learning

  • Learn from reward and consequence
  • Powers game-players and chat-model tuning
  • The agent generates its own training signal
  • Human parallel: a child learning by trial, reward, and consequence
The agent acts, the grid returns state + reward — tiles rise toward the reward

Unifying View

  • Self-supervised prediction builds the world model
  • Reinforcement fine-tunes on top
  • The same recipe in human brain and in models

III. “AI”: An Ambiguous Term

Beginning: Dartmouth, 1956

  • McCarthy chose “artificial intelligence”
  • The conjecture: intelligence can be “so precisely described that a machine can be made to simulate it”
  • They thought a summer and 10 assistants would do it.
McCarthy
Minsky
Shannon
with Nathaniel Rochester — the four
who signed the 1955 proposal

Four Acronyms

  • ML — algorithms that improve at a task with more data
  • ANI — narrow: superhuman at one task, useless outside it
  • AGI — general: human-level breadth, novel tasks included
  • ASI — beyond: exceeds the best humans at essentially everything

ANI is here · AGI is the aspiration · ASI is the artwork

Make Coffee

  • Functionalist test: a machine does what a human did — intelligent?
  • Once we see the mechanism, we stop calling it AI
  • Chess, OCR, translation — “AI” until it worked
  • Make coffee in any kitchen in the world
  • “AI” often means: not yet understood

Coffee, yes — but in a rig built for it, not any kitchen

Turing Test

  • Imitation: indistinguishable in text → passes
  • Says nothing about internal states
  • Cited often, engineered against rarely
  • A floor, not a ceiling

Russell & Norvig: Agents

  • Act to achieve the best expected outcome
  • The textbook definition — agent, environment, measure
  • Says nothing about how — only that it decides well
Agent Environment action percepts utility / performance measure what it maximizes

Chollet: Intelligence as Adaptation

  • Efficiency at acquiring new skills
  • Adaptation, not static performance
  • The machine echo of human skill acquisition
  • Memorize well, adapt poorly → not intelligent

One Definition of Knowledge

  • Data — raw, uncompressed
  • Compression — redundancy found
  • Abstraction — the reusable structure it yields
  • Prediction — the test on unseen data
  • Knowledge — abstraction that keeps passing

Process data to generate knowledge

Knowledge as Compression

Three traditional sources of knowledge:

  • Genes — survival encoded by evolution
  • Experience — direct interaction
  • Culture — collective wisdom, transmitted

A fourth new source, over the last 50 years:

  • Computers — synthesis at scale
Compression is finding the structure — 3 numbers per point, 2 that matter

IV. Generative AI & LLMs

What Gen AI Is

  • Simulation from a partially learned distribution; produces new samples
  • Text, image, audio, code
  • Self-supervised pre-training + the transformer
  • Planetary scale recursive classification simulation
Noise → sample: a diffusion model inventing characters

What Gen AI Is Not

  • No guarantee of truth, cause, or grounding
  • Fluency is not knowledge
  • It compresses culture and lets us query it
  • New and powerful — not a mind

The LLM

  • A next-token predictor over human text
  • No persistent memory, weak causal models
  • Unreliable deliberate reasoning
  • A statistical reflection of us
Attention, layer by layer — every token weighing every other

V. The Five Tribes

Symbolists — Logic and Deduction

  • Roots: logic — Newell & Simon’s symbol system hypothesis
  • Learning = inverse deduction; rules, trees, expert systems
  • Precise and auditable — but brittle
  • Optimization: combinatorial search over rule/hypothesis space
  • Modern echo: neuro-symbolic AI, formal verification
  • EXAMPLE: Routing Email Complaints

Every answer comes with its own audit trail

Josh Tenenbaum

Connectionists — Brain Inspired

  • Roots: McCulloch–Pitts (1943), Rosenblatt’s perceptron (1958)
  • Backpropagation (1986) → CNNs → transformers
  • Optimization: stochastic gradient descent via backpropagation
  • Scales with data and compute — and dominates today
  • EXAMPLE: Making DNA viral capsids
Signals flowing forward, layer by layer — the 1943 cartoon, stacked
Hinton
LeCun
Bengio
Schmidhuber

Connectionists: Bolivarian Contribution

  • Two inspiring contributors from the Bolivarian nations!

Evolutionaries — Selection Algorithm

  • Roots: Darwin, formalized by Holland (1975)
  • Evolve a population — mutation, crossover, selection
  • No gradient, no labels — only a fitness score
  • Optimization: genetic algorithms / evolutionary search
  • Niche but creative: architecture search, robot morphology
  • EXAMPLE: Avoiding immune response for gene envelope
John H. Holland
Selection on a fitness landscape — a population climbs, some get stuck

Bayesians — Belief as Probability

  • Roots: Bayes (1763), Laplace
  • Learning = updating a distribution over hypotheses
  • Prior + likelihood → posterior
  • Optimization: probabilistic inference (MCMC, variational inference)
  • Calibrated uncertainty, at a computational cost
  • EXAMPLE: Soldier Death Benefit Pensions

\[\underset{\color{#f0a868}{\text{posterior}}}{\color{#f0a868}{P(H \mid D)}} \;=\; \frac{\underset{\color{#63b3a6}{\text{likelihood}}}{\color{#63b3a6}{P(D \mid H)}}\;\cdot\;\underset{\color{#7aa0e0}{\text{prior}}}{\color{#7aa0e0}{P(H)}}}{\underset{\color{#9aa0aa}{\text{evidence}}}{\color{#9aa0aa}{P(D)}}}\]

\[\underset{\color{#e06c75}{\text{predictive}}}{\color{#e06c75}{P(x^{*} \mid D)}} \;=\; \int \color{#63b3a6}{P(x^{*} \mid H)}\; \color{#f0a868}{P(H \mid D)}\; dH \qquad \text{— average over every hypothesis, don't pick one}\]

Analogizers — Similarity

  • Roots: psychology — reasoning from prior cases
  • k-nearest neighbors, SVMs, the kernel trick
  • Optimization: constrained optimization to maximize margin (SVMs)
  • Legacy: recommendation, few-shot via embeddings
  • EXAMPLE: Locating outperforming stock pairs

Vladimir Vapnik — co-inventor of the SVM
The kernel trick — lift to a space where a flat plane separates them

Breiman: Three Received Wisdoms

Leo Breiman lecturing

Leo Breiman, Statistical Modeling: The Two Cultures (2001)

Name Received Wisdom Breiman’s Point Neural Nets
Three men sheltering under the gate in Kurosawa's film RashomonRashomon There is one true model Many models fit about equally well Small net, 100 runs: 32 minima, about equal test error
William of Ockham in a stained glass windowOccam Simpler is better Simpler often costs accuracy Usually beat linear regression, far harder to read
Richard Bellman holding a computer printoutBellman Many variables are a curse Many variables can help Raw pixels in, features learned

Rashomon and Occam nets: Breiman’s examples. Bellman net: my reading.

The Frontier Borrows From All Five

  • No production system is one tribe
  • LLMs: connectionist core, Bayesian uncertainty, symbolic tools, RL tuning
  • Domingos’ provocation: one master algorithm?
  • Which synthesis will your work advance?

VI. State of the Art

Brittle AI & the Reasoning Gap

  • Narrow, brittle outside the training distribution
  • The frontier problem is sample-efficient reasoning, not scale
  • Performance can mask absent understanding
  • ARC-AGI-3 saturated in a year — at thousands of dollars in compute
  • The compute-capped ARC Prize is still unclaimed

System 1 vs. System 2

  • After Kahneman: fast intuition, slow deliberation
  • System 1 — crystallized knowledge: answers in one fast pass of inference
  • System 2 — learning: slow, effortful, grows what isn’t there yet
  • Practice turns System 2 into System 1
  • LLMs: a vast crystal, frozen after training — “reasoning” isn’t yet learning
Daniel Kahneman
Familiar questions flash through the crystal; a new one makes it grow

Reality Gets a Vote

  • Knowledge is what survives a test
  • Code and maths moved fastest — they can be checked
  • Fluency with no verifier → hallucination
  • Essays, strategy, taste, most real tasks: no oracle

Mike Tyson standing over a floored opponent, captioned "Everyone has a plan 'till they get punched in the mouth."

Pearl’s Ladder

  • See — P(y | x)
  • Do — P(y | do(x))
  • Imagine — what if I had?
  • LLMs just see, no doing or imagining without a verifier!
  • Can’t watch your way to cause

The Experiment Is the Intervention

  • do(x) — reach in, set x by hand, cutting its usual causes
  • That severed graph is what an experiment is — randomise, don’t just watch
  • Verifying and causing are one move
See · P(slippery | sprinkler)SeasonRainSprinklerGrass wetGrass slipperyGardenerDo, at random · P(slippery | do(sprinkler))✕SeasonRainSprinklerGrass wetGrass slipperyRandomization

What “Off” Tells You

\[\underset{\color{#f0a868}{\text{posterior}}}{\color{#f0a868}{P(\text{monsoon} \mid \text{off})}} \;=\; \frac{\underset{\color{#63b3a6}{\text{likelihood}}}{\color{#63b3a6}{P(\text{off} \mid \text{monsoon})}}\;\cdot\;\underset{\color{#7aa0e0}{\text{prior}}}{\color{#7aa0e0}{P(\text{monsoon})}}}{\underset{\color{#9aa0aa}{\text{evidence}}}{\color{#9aa0aa}{P(\text{off})}}}\]

See the gardener decides, by the season

Do at random: randomization decides

area = probability
dry season ⅔on ¾off ¼rain 3%monsoon ⅓on ¼off ¾rain 67%dry season ⅔on ½off ½rain 3%monsoon ⅓on ½off ½rain 67%
prior P(monsoon)
⅓ Jun–Sep: 122 of 365 days
⅓ the same calendar
likelihood P(off | season)
¾ monsoon, ¼ dry the gardener · ratio 3
½ monsoon, ½ dry randomized · ratio 1
evidence P(off)
P(off) = Σs P(off | s)·P(s)
= ¾·⅓ + ¼·⅔ = 5⁄12
P(do(off)) = Σs Prandom(off | s)·P(s)
= ½·⅓ + ½·⅔ = ½
posterior P(monsoon | off)
¾·⅓ ÷ 5⁄12 = ⅗ “off” lifts ⅓ to ⅗
½·⅓ ÷ ½ = ⅓ the prior again
so, slippery P(slippery | off)
P(slippery | off) = Σs P(slippery | s)·P(s | off)
= 67%·⅗ + 3%·⅖ = 41%
P(slippery | do(off)) = Σs P(slippery | s)·P(s | do(off))
= 67%·⅓ + 3%·⅔ = 24%
  • off ≠ do(off): different distributions over the season, P(monsoon | off) = ⅗ vs P(monsoon | do(off)) = ⅓. Same P(slippery | s) inside each season; only the weights differ
  • Randomization’s likelihood ratio is 1: posterior = prior, even for unrecorded causes

Why Randomization Works

Every Mumbai day, 1961–90: the true rates

dry seasonOct–May monsoonJun–Sep
share of the year calendar ⅔ ⅓
it rains real record 3% 67%
gardener waters assumed 75% 25%
slippery if sprinkler on 100% 100%
slippery if off 3% 67%
sprinkler’s effect +97 pts +33 pts

See season not recorded

P(monsoon | off) = ⅗, by Bayes

100% − 41% = +59 pts ✗

thirty years of data didn’t fix it

See + record the season adjust

compare inside each season, weight by its share

⅔·97 + ⅓·33 = +76 pts ✓

only for causes you knew to record

Do at random

P(monsoon | do(off)) = ⅓, the prior

100% − 24% = +76 pts ✓

balances causes nobody recorded

see  ∑s P(y | x, s) · P(s | x)    do  ∑s P(y | x, s) · P(s)    seeing weights by the posterior, doing by the prior

  • Adjusting fixes the confounders you know; randomization fixes them all
  • To do experiments and find cause, you need a body

VII. What We Want from AGI

Generality, Efficiency, Reasoning

  • Generality — competence transfers to the untrained
  • Sample Efficiency — human-like, from few examples
  • Robust Reasoning — multi-step, causal, self-correcting

Memory, Grounding, Aligned Autonomy

  • Persistent Memory — no catastrophic forgetting
  • Grounded World Models — causal, ideally embodied
  • Autonomy + Alignment — open-ended yet controllable
  • The last is a constraint — and the highest-stakes one

VIII. The Future

Foundation Models Meet Robotics

  • Model “brains” in capable “bodies”
  • Surgery, deep-sea exploration, autonomous logistics
  • The loop closes: sense → predict → act → learn
  • Where the next decade concentrates

Unitree G1 — a body waiting for a better brain

AI Ethics — Beyond My Expertise

  • Cheap Dopamine — engineered attention capture
  • Economic Effects — labour, wealth concentration
  • Children — growing up AI-saturated
  • Distribution — who reaps the fruits of the technology
  • A prior, held lightly: fewer regulations tend to serve better

Key Takeaways

  • “AI” is an ambiguous term — today it means LLMs
  • Minds run on prediction — octopus to cortex to transformer
  • Human intelligence is the ML task list, written first in our biology
  • The open gaps — reasoning, memory, grounding, alignment — are the agenda

Thank You

Questions?

QR code linking to these slides at tinyurl.com/AI-past

tinyurl.com/AI-past
project-delphi.github.io/ai-past-present-future-talk

Key References I

Foundations & cognition

  • Turing, A. (1950). Computing Machinery and Intelligence. Mind.
  • Russell, S. & Norvig, P. (2020). Artificial Intelligence: A Modern Approach (4th ed.)
  • Kahneman, D. (2011). Thinking, Fast and Slow.
  • Friston, K. (2010). The free-energy principle: a unified brain theory? Nature Reviews Neuroscience.
  • Murray, J. et al. (2014). A hierarchy of intrinsic timescales across primate cortex. Nature Neuroscience.
  • Northoff, G. & Huang, Z. (2017). How do the brain’s time and space mediate consciousness and its different dimensions? Neurosci. Biobehav. Rev.

Key References II

Symbolic & neural roots

  • Newell, A. & Simon, H. (1976). Computer science as empirical inquiry: symbols and search. CACM.
  • McCulloch, W. & Pitts, W. (1943). A logical calculus of ideas immanent in nervous activity.
  • Rosenblatt, F. (1958). The perceptron. Psychological Review.
  • Minsky, M. & Papert, S. (1969). Perceptrons.

Key References III

Learning at scale

  • Rumelhart, D., Hinton, G. & Williams, R. (1986). Learning representations by back-propagating errors. Nature.
  • Vaswani, A. et al. (2017). Attention Is All You Need. NeurIPS.
  • Hinton, G., Bengio, Y. & LeCun, Y. — 2018 ACM A.M. Turing Award, for deep neural networks
  • Dean, J. et al. (2012). Large scale distributed deep networks. NeurIPS.

Key References IV

Defining intelligence

  • Holland, J. (1975). Adaptation in Natural and Artificial Systems.
  • Sutton, R. (2019). The Bitter Lesson.
  • Chollet, F. (2019). On the measure of intelligence. arXiv:1911.01547.
  • Domingos, P. (2015). The Master Algorithm.

Key References V

Models, causality & data

  • Breiman, L. (2001). Statistical modeling: the two cultures. Statistical Science.
  • Rudin, C. (2025). Leo Breiman, the Rashomon effect, and the Occam dilemma. arXiv:2507.03884.
  • Pearl, J. & Mackenzie, D. (2018). The Book of Why.
  • Menne, M. et al. (2012). Global Historical Climatology Network – Daily (GHCN-Daily), v3. NOAA NCEI. Station IN012070800, Bombay/Santacruz.

Image Credits

  • Bonobo — natataek, CC BY-SA 3.0
  • Octopus — Diego Delso, CC BY-SA 4.0
  • L. Messi goal clip (Getafe, 2007) — broadcast footage, rights holder unidentified
  • Mirror self-recognition (bear, infant, child animations) — via Medium
  • G. Hinton — Cmichel67, CC BY-SA 4.0
  • Y. LeCun — J. Barande, CC BY-SA 2.0
  • Y. Bengio — Xuthoria, CC BY-SA 4.0
  • J. Schmidhuber — ITU/R. Farrell, CC BY 2.0
  • D. Kahneman — nrkbeta, CC BY-SA 2.0
  • J. Pearl — Better Than Bacon, CC BY 2.0
  • J. McCarthy — null0, CC BY-SA 2.0
  • M. Minsky — OLPC, CC BY 3.0
  • C. Shannon — Tekniska museet, CC BY 2.0
  • F. Chollet — Ramosset, CC BY-SA 4.0
  • A. Turing — Elliott & Fry, public domain
  • T. Bayes — public domain (attribution disputed)
  • Soviet flag — public domain
  • 3D visuals — three.js (mrdoob and contributors), MIT
  • QR codes (tinyurl.com/AI-past, link to these slides) — generated with segno, no rights claimed
  • Early human migrations map — Dbachmann, CC BY-SA 4.0
  • Acheulean handaxe (Saint-Acheul) — Didier Descouens, CC BY-SA 4.0
  • Map boundaries — Natural Earth, public domain
  • J. Holland, V. Vapnik, J. Tenenbaum, K. Friston — no free licence identified; used under fair dealing for education
  • Portrait with sculpture — personal photograph, used with permission
  • Attention animation — BertViz, via comet.com
  • Diffusion denoising animation — via Towards Data Science
  • Decision tree animation — via Medium
  • ANI/AGI/ASI illustration — via Medium
  • AI history timeline — via LinkedIn
  • Unitree G1 robot animation — via roboticgizmos.com
  • Coffee-making robot animation — via Mothership
  • ARC-AGI task animation & ARC-AGI-3 leaderboard — © ARC Prize Foundation, via arcprize.org
  • M. Tyson quote poster — rights holder unidentified; used under fair dealing for education
  • William of Ockham (stained glass, Surrey; cropped) — Moscarlop, CC BY-SA 3.0
  • Rashomon (1950) press photo — Daiei Studios; tagged public domain on Commons (US, not renewed), status not independently confirmed
  • L. Breiman (SAMSI, via Wikipedia), R. Bellman (© IEEE, via ETHW) — no free licence identified; used under fair dealing for education

Photographs from Wikimedia Commons unless noted. Titled slide image: Piedras del Tunjo.

Discussion

What is the most important open problem in AI — and how will you spend the next five years on it?