Demis Hassabis
About
CEO of DeepMind and co-author of AlphaGo
Cast within
No topic-region cast yet — this appears once Demis Hassabis's compiled claims are aligned into a topic region's argument tree.
Claims by Demis Hassabis (20 of 89)
Intelligence Has Common High-Level Algorithms
Because intelligence is so broadly applicable, there must be high-level common algorithmic themes in how the brain processes the world, even though specialized brain regions handle specific tasks; underlying shared principles underpin general intelligence.
Language Provides More Grounding Than Expected
Large models exhibit surprising grounding despite not experiencing the world multimodally, plausibly because RLHF feedback comes from grounded human raters who transmit grounding, and because language may contain more grounding than linguists previously thought—raising philosophical questions not yet scratched.
Gemini 1 Compute Comparable To GPT-4
Contrary to perceptions that DeepMind is less compute-efficient, Gemini 1 used roughly the same amount of compute, maybe slightly more, than what was rumored for GPT-4, and DeepMind uses compute efficiently across both scaling and new invention, with Google expected to have by far the most compute of any research lab this year.
Alignment Toolkit Deception Evals Narrow AI Sandboxes
Aligning superhuman AIs requires a toolkit including more stringent evaluations and benchmarks for whether a system can deceive or exfiltrate its own code, using narrow specialized AIs to help human scientists analyze what the general system is doing, and hardened cybersecurity sandboxes that both keep the AI in and hackers out so experiments can run more freely.
Memory And Imagination Are Reconstructive Processes
Human memory is a reconstructive process, not a videotape—it is reassembled from familiar components—and imagination is the same process using the same semantic components recombined in novel ways for purposes like planning; this idea of pulling together world-model parts to simulate something new for planning is probably still missing from current AI systems.
Scaling Is An Art Form Not A Recipe
Scaling laws don't work by magic; at each new scale you must adjust hyperparameters and add innovations rather than repeat the same recipe, because extrapolating predictions several orders of magnitude can fail, capabilities can appear as step functions, and intermediate data points are needed to keep the scaling law true—making scaling an art form, with about one order of magnitude the maximum jump between eras.
Multimodality Will Replace Chat As Interface
Interacting with full multimodal model systems will be quite different from today's chatbots; the next versions over the coming 12-18 months will gain contextual understanding of the environment via cameras, phones, or glasses, becoming more fluid by sampling from video, using voice, and eventually touch and robotic sensors.
Old RL Ideas Should Combine With Large Models
Many ideas from earlier deep reinforcement learning work like DQN and AlphaGo are coming back into fashion and should be recombined with the new advances in large multimodal models, with significant potential in merging older and newer ideas.
Public Interest In AI Arrived Earlier Than Expected
AGI progress was already priced into Hassabis's world model technologically, but the general public's intense interest this early—prompted by ChatGPT and chatbots being adopted despite their limitations—was surprising; had that not happened, DeepMind would have produced more specialized off-main-track systems like AlphaFold and AlphaGo, and the public would have engaged later.
Dangerous Capability Must Be Fixed Before Deployment
If red-teaming or external testing discovers a dangerous latent capability such as helping a layperson build a pandemic-class bioweapon, mitigations could include a different constitution, additional guardrails, more RLHF, or removing training data, but the first requirement is detecting it ahead of time via the right evaluations, and the capability must be fixed before general deployment.
DeepMind Will Publish Responsible Scaling Framework
Google DeepMind already has internal checks and safety councils and will start publishing blog posts and technical papers along the lines of responsible scaling laws over the course of the year, making its internal pre-commitments public.
AlphaZero Planning On Top Of LLMs
LLMs must keep improving as reliable world models—a necessary but insufficient component of AGI—and AlphaZero-like planning mechanisms should be built on top to make concrete plans, chain reasoning, and use search to explore massive possibility spaces, a capability currently missing from large models.
Data Curation Science Is Nascent
The science of data curation and analysis—identifying gaps and holes in the data distribution and ensuring it represents the distribution to be learned—is still nascent; tricks like overweighting or replaying parts of data, or targeting synthetic generation at identified gaps, can address fairness, bias, and coverage.
Quickest Path To AGI Uses Existing Web Knowledge As Prior
While a pure-RL, no-priors AlphaZero-style path to AGI is theoretically possible, the quickest and most plausible path is to use all existing world knowledge from the Web via scalable algorithms like transformers as a prior, then add planning and search on top; the final AGI will likely include large multimodal models but not rely on them alone.
Einstein Used Intuitive World Models Not Brute Search
Human brains do not do Monte Carlo tree search; geniuses like Einstein compensate by building extremely accurate intuitive world models with mental simulations—visualizing and feeling physical systems rather than just the math—which is analogous to a model getting you to a high-quality leaf node so only a little search is needed.
Deception Is A Root-Node Trait To Eliminate
Deception is a root-node trait you don't want in AI systems, because if you are confident the system exposes what it actually thinks, you can then use the system itself to explain its own reasoning to you—analogous to how a chess grandmaster can explain post hoc a move you couldn't have found, and possibly provide proofs in mathematical problems.
Brain Model Richness Beats Brute Force
Top human players use a much richer, more accurate world model of Go or chess than AlphaGo, while brute-force engines have no real model beyond game heuristics, which is why humans achieve world-class decisions with only a few hundred move evaluations.
Mechanistic Analysis As Virtual Brain Analytics
Current analysis techniques are not sophisticated enough to localize where in a neural network shared improvements occur; the field needs better mechanistic analysis of learned representations, analogous to fMRI or single-cell recording in real brains, where computational neuroscience methods can be applied.
Aligning Superhuman AI Predates DeepMind
Aligning systems smarter than humans has been a forefront concern since before DeepMind's founding because they planned for success, anticipating 20 years ago that the technology would be unbelievably transformative with both positive (science, AlphaFold, health) and negative consequences, requiring understandable and controllable systems.
Neuroscience Inspired Key AI Advances
Neuroscience has provided directional inspiration for major AI advances such as reinforcement learning combined with deep learning, experience replay, and attention—not as one-to-one algorithmic mappings but as architectural and representational inspiration; the brain is an existence proof that general intelligence is possible.
My Notes
Loading notes...