📘 1.1.3 — AI winters and milestones
Topic: Foundations of artificial intelligence
AI history is not a steady climb from primitive rules to inevitable intelligence. It is a sequence of useful ideas, impressive demonstrations and difficult tests of whether those demonstrations scale.
The 1956 Dartmouth workshop helped establish artificial intelligence as a research field. Early programs proved theorems and played games; these were genuine milestones. But success on a small, carefully structured task did not establish that the same method could handle an open-ended world.
One obstacle was combinatorial explosion. Searching through possible action sequences can work in a tiny puzzle and become impractical when the number of choices or the solution length grows. Faster hardware helps, but does not erase the underlying growth. The 1973 Lighthill report criticized this failure to scale and contributed to major reductions in British AI research support.
Another obstacle concerned representation. Minsky and Papert’s 1969 analysis exposed limits of simple perceptrons. A single linear threshold unit cannot express XOR: it cannot separate the two inputs that differ from the two that agree. That limitation does not prove that multilayer neural networks are incapable of learning; confusing a restricted model with an entire research direction is the wrong inference.
Expert systems offered another route: encode specialist knowledge rather than rely on general search alone. DENDRAL used chemistry knowledge to narrow candidate molecular structures. Commercial expert systems then expanded rapidly. Yet complex systems were expensive to build and maintain, and their handling of uncertainty and inability to learn from experience created serious weaknesses. After the late-1980s boom, many businesses failed to meet their promises—the period commonly called an AI winter.
The recovery involved several changes, not one magical invention. Neural-network research revived around backpropagation; probabilistic reasoning and machine learning gained influence; larger datasets and more computing supported later advances. The textbook highlights the 2012 ImageNet breakthrough as a major deep-learning milestone. The 2015 Nature review explains the central mechanism: multiple processing layers learn increasingly abstract representations.
There is an important computational detail behind that last sentence. Stanford’s neural-network notes show why nonlinear activations matter: stacking linear transformations alone still gives a linear transformation. More layers are useful because of what they compute and learn, not simply because a network is drawn deeper.
Reading check: when a system works in a demonstration, ask three questions. What can its representation express? How does its computation grow with problem size? What evidence shows that it works beyond the examples that made it famous? These are practical lessons from AI winters, not a prediction of when another one will occur.
Topic: Foundations of artificial intelligence
AI history is not a steady climb from primitive rules to inevitable intelligence. It is a sequence of useful ideas, impressive demonstrations and difficult tests of whether those demonstrations scale.
The 1956 Dartmouth workshop helped establish artificial intelligence as a research field. Early programs proved theorems and played games; these were genuine milestones. But success on a small, carefully structured task did not establish that the same method could handle an open-ended world.
One obstacle was combinatorial explosion. Searching through possible action sequences can work in a tiny puzzle and become impractical when the number of choices or the solution length grows. Faster hardware helps, but does not erase the underlying growth. The 1973 Lighthill report criticized this failure to scale and contributed to major reductions in British AI research support.
Another obstacle concerned representation. Minsky and Papert’s 1969 analysis exposed limits of simple perceptrons. A single linear threshold unit cannot express XOR: it cannot separate the two inputs that differ from the two that agree. That limitation does not prove that multilayer neural networks are incapable of learning; confusing a restricted model with an entire research direction is the wrong inference.
Expert systems offered another route: encode specialist knowledge rather than rely on general search alone. DENDRAL used chemistry knowledge to narrow candidate molecular structures. Commercial expert systems then expanded rapidly. Yet complex systems were expensive to build and maintain, and their handling of uncertainty and inability to learn from experience created serious weaknesses. After the late-1980s boom, many businesses failed to meet their promises—the period commonly called an AI winter.
The recovery involved several changes, not one magical invention. Neural-network research revived around backpropagation; probabilistic reasoning and machine learning gained influence; larger datasets and more computing supported later advances. The textbook highlights the 2012 ImageNet breakthrough as a major deep-learning milestone. The 2015 Nature review explains the central mechanism: multiple processing layers learn increasingly abstract representations.
There is an important computational detail behind that last sentence. Stanford’s neural-network notes show why nonlinear activations matter: stacking linear transformations alone still gives a linear transformation. More layers are useful because of what they compute and learn, not simply because a network is drawn deeper.
Reading check: when a system works in a demonstration, ask three questions. What can its representation express? How does its computation grow with problem size? What evidence shows that it works beyond the examples that made it famous? These are practical lessons from AI winters, not a prediction of when another one will occur.
Russell & Norvig — AIMA, 4e, §1.3
Stanford CS231n — Neural networks
LeCun, Bengio & Hinton — Nature (2015)
A complete search tree has branching factor 3 and root depth 0. How many nodes are there in total from depth 0 through depth 4, including the root?
Anonymous Quiz
0%
81
100%
120
0%
121
0%
243
A classifier outputs 1 iff w1*x1 + w2*x2 + b > 0. Its binary inputs must produce XOR: 00→0, 10→1, 01→1, 11→0. Which statement is correct?
Anonymous Quiz
0%
Some real weights and bias implement all four labels
0%
No real weights and bias implement all four labels
100%
Setting both weights to zero implements XOR
0%
Removing the bias always implements XOR
🆕 Gemini 4 Carbon — a restricted first rollout
Company-reported release facts and changes:
Google announced Gemini 4 Carbon on 28 October, building on Gemini 4 Argon. The company emphasizes longer engineering workflows, coding and cybersecurity defense. It reports improvements over Argon in secure-code generation and fewer unsupported claims in internal testing. These are company-reported findings, not guarantees for every deployment.
Access is initially limited: Google says it will begin with a subset of cyber defenders and expand to trusted partners through its Fairwind Program. Broader developer, business and consumer availability is planned rather than established by this announcement.
Artificial Analysis already lists Carbon (high) among its leading models. That does not establish a universal winner or an apples-to-apples improvement over Argon. The matched predecessor comparison below remains pending verification of exact scores and evaluation conditions.
Requested Artificial Analysis Intelligence Index matched Carbon–Argon comparison (v5.0; points); status as of 2026-10-29: independent results pending; no ranking.
Limitations: Pending means this exact matched predecessor comparison is not yet verified, NOT that independent evaluations do not exist. The retrieved page lists Carbon (high), but exact paired scores, model variants, test dates and full evaluation setup were not extracted. No invented scores, delta or universal ranking.
Company-reported release facts and changes:
Google announced Gemini 4 Carbon on 28 October, building on Gemini 4 Argon. The company emphasizes longer engineering workflows, coding and cybersecurity defense. It reports improvements over Argon in secure-code generation and fewer unsupported claims in internal testing. These are company-reported findings, not guarantees for every deployment.
Access is initially limited: Google says it will begin with a subset of cyber defenders and expand to trusted partners through its Fairwind Program. Broader developer, business and consumer availability is planned rather than established by this announcement.
Artificial Analysis already lists Carbon (high) among its leading models. That does not establish a universal winner or an apples-to-apples improvement over Argon. The matched predecessor comparison below remains pending verification of exact scores and evaluation conditions.
Requested Artificial Analysis Intelligence Index matched Carbon–Argon comparison (v5.0; points); status as of 2026-10-29: independent results pending; no ranking.
Limitations: Pending means this exact matched predecessor comparison is not yet verified, NOT that independent evaluations do not exist. The retrieved page lists Carbon (high), but exact paired scores, model variants, test dates and full evaluation setup were not extracted. No invented scores, delta or universal ranking.
Google DeepMind — Gemini 4 Carbon
Artificial Analysis — model comparison
In a factory, 1% of parts are defective. A scanner flags 90% of defective parts and 10% of nondefective parts. A randomly selected part is flagged. What is the probability that it is defective?
Anonymous Quiz
50%
1%
50%
8.33% (approximately)
0%
90%
0%
98.90% (approximately)
A* has four frontier paths ending at A, B, C and D. Their (g, h) values are A=(2,8), B=(6,1), C=(4,4), D=(9,0), where g is cost so far and h estimates remaining cost. Which path does standard A* select next?
Anonymous Quiz
0%
The path to A
100%
The path to B
0%
The path to C
0%
The path to D
Discrete random variables X and Y are conditionally independent given Z. Which statement must hold for every z with P(Z=z)>0?
Anonymous Quiz
0%
X and Y are also independent without conditioning on Z.
50%
P(X,Y | Z=z) = P(X | Z=z) P(Y | Z=z).
50%
X and Y cannot both equal 1 given Z=z.
0%
The distribution of X does not depend on Z.
A regression model must predict the same real number c for three equally weighted training targets: 1, 2 and 9. Which c minimizes the mean squared error on these three examples?
Anonymous Quiz
0%
2
0%
3
100%
4
0%
5
🧠 Microsoft’s research roundup puts code understanding in focus
Microsoft Research’s 29 October roundup highlights a stubborn problem for coding agents: understanding unfamiliar code, not simply changing it.
The episode describes CABRA, a framework that generates coding challenges with controlled complexity to help isolate where agents struggle. Microsoft reports that tools helped agents overcome some programming obstacles, while understanding unfamiliar codebases remained difficult.
Another featured project, TeleTune, explores learning reusable computer-use skills from activity logs. Such logs record actions, but do not necessarily reveal the user’s goals or separate one task from another.
These are distinct research projects summarized by Microsoft—not a single new agent release or independent proof that the approaches will generalize to every real-world workflow. The useful distinction is between making an agent act and establishing that it understands the task it is acting on.
Microsoft Research
Microsoft Research’s 29 October roundup highlights a stubborn problem for coding agents: understanding unfamiliar code, not simply changing it.
The episode describes CABRA, a framework that generates coding challenges with controlled complexity to help isolate where agents struggle. Microsoft reports that tools helped agents overcome some programming obstacles, while understanding unfamiliar codebases remained difficult.
Another featured project, TeleTune, explores learning reusable computer-use skills from activity logs. Such logs record actions, but do not necessarily reveal the user’s goals or separate one task from another.
These are distinct research projects summarized by Microsoft—not a single new agent release or independent proof that the approaches will generalize to every real-world workflow. The useful distinction is between making an agent act and establishing that it understands the task it is acting on.
Microsoft Research
A CSP has D(X)={1,2,3,4}, D(Y)={2,3}, and the only constraint X<Y. After revising only the directed arc X→Y to make X arc-consistent with Y, what is D(X)?
Anonymous Quiz
0%
{1}
33%
{1,2}
33%
{1,2,3}
33%
{1,2,3,4}
Looking to get acquainted with neuroscience? Join a friendly space to learn the basics, share resources, and explore the field together.
👉 Join here: https://t.me/neuroprocessing_community
📩 Share this with anyone who might be interested.
👉 Join here: https://t.me/neuroprocessing_community
📩 Share this with anyone who might be interested.
🧬 AI searches human proteins for mRNA-delivery scaffolds
Researchers reporting in Nature Communications have developed DeepDelivery, an AI-guided platform for finding human proteins that could help carry mRNA into cells.
The system nominated 512 candidates from the human proteome. In experiments on 15 selected proteins, 12 supported particle formation and mRNA delivery within the researchers’ modular engineering framework. Several belonged to the TRIM family, better known for roles in antiviral responses.
The team also used model interpretation and perturbation analyses to guide changes to TRIM25, improving delivery in their experimental system.
The important boundary is cultured cells, not demonstrated treatment in people. These results identify promising scaffold components; they do not establish that a human-derived carrier is automatically safe, avoids immune reactions, or is ready for clinical use.
Nature Communications
Researchers reporting in Nature Communications have developed DeepDelivery, an AI-guided platform for finding human proteins that could help carry mRNA into cells.
The system nominated 512 candidates from the human proteome. In experiments on 15 selected proteins, 12 supported particle formation and mRNA delivery within the researchers’ modular engineering framework. Several belonged to the TRIM family, better known for roles in antiviral responses.
The team also used model interpretation and perturbation analyses to guide changes to TRIM25, improving delivery in their experimental system.
The important boundary is cultured cells, not demonstrated treatment in people. These results identify promising scaffold components; they do not establish that a human-derived carrier is automatically safe, avoids immune reactions, or is ready for clinical use.
Nature Communications
❤1
Choose the action with maximum expected utility. A: utility 12 with probability 0.25, otherwise 0. B: utility 4 for certain. C: utility 10 with probability 0.5, otherwise −4. D: utility 30 with probability 0.1, otherwise −2. Which action is optimal?
Anonymous Quiz
0%
A
50%
B
50%
C
0%
D
In the standard agent model, what does an agent function map to an action?
Anonymous Quiz
0%
Only the most recent percept
50%
The complete percept sequence received so far
0%
The source code of the agent program
50%
The environment’s true state, including unobserved information
❤2