Mechanism◆Video · 42:17 · 2m
Hassani describes AFMD, Liquid's automated foundation model design system, which uses an evolution strategy with real hardware and downstream-task evaluation to eliminate human bias from architecture choices — a deliberate reaction against how architecture decisions are typically made ad hoc by a small group of experts at frontier labs.
Mechanism◆Audio · 14:08 · 5m
Scaringe contrasts the traditional domain-based architecture (100-150 separate computers, each running supplier-written software on its own island) with Rivian's zonal architecture (1-3 computers running a single OS). He cites Rivian's $5.8 billion deal with Volkswagen Group as evidence that legacy automakers recognize they cannot build this capability internally.
ContextAudio · 6:18 · 1m
Feldman explains the strategic thesis behind Cerebras: every major new compute workload — graphics, mobile — created a new dominant architecture while incumbents like Intel and AMD got zero share. AI represented the same kind of opportunity, demanding a clean-sheet design, not a derivative of existing architectures.
MechanismVideo · 43:09 · 2m
Hassani explains that Liquid built its AFMD automated architecture-search system with hardware-in-the-loop evolutionary optimization specifically to remove the human biases he says still drive architecture decisions even at top labs like Anthropic and OpenAI.
Claim◆Video · 12:12 · 2m
Jensen argues that a computing architecture's install base is the decisive factor for its success, using x86's triumph over more elegant RISC architectures as evidence.
ClaimArticle · 53 words
Raschka observes that 2026 architecture innovation spans several new directions—hybrid designs, state space layers, MoE capacity routing, and more—rather than just building bigger transformers.
MechanismAudio · 3:15 · 1m
Feldman explains Cerebras's bet on wafer-scale chips: to be 15-20x faster than GPUs, you cannot make minor modifications — you need a fundamentally different architecture, which is why Cerebras built a dinner-plate-sized chip.
ClaimVideo · 53:41 · 2m
Hasani traces the input-dependent gating mechanism now central to architectures like Mamba back to Liquid AI's own Liquid-S4 paper, published a year and a half earlier, framing it as a foundational discovery rooted in liquid neural network theory.
DataVideo · 14:51 · 2m
When Osmo visualized their learned odor embedding space, semantically related scent categories clustered into nested neighborhoods (florals containing jasmine/rose/violet) that the model was never told to form.
AnecdoteVideo · 4:42 · 3m
Dianne describes how shipping Golden Gate Claude — a quirky 24-hour experiment where Claude obsessed over the Golden Gate Bridge — was an early inflection point that showed Anthropic could build products different from competitors.
Claim◆Video · 16:27 · 16m
In Wyart's synthetic tree-structured world, a shallow network does exactly what Chomsky predicted — it memorizes and cannot generalize. But deep architectures exhibit a strong implicit bias to construct coarse-grained hierarchical variables, learning to be creative from only polynomially many sentences. This is a counterexample showing that what must be 'innate' shrinks dramatically for a deep architecture.
MechanismVideo · 29:50 · 2m
Hasani explains that the core bottleneck in scaling architectures like liquid neural networks is that nonlinear relationships resist being cleanly tensorized for parallel computation, which is why state space models default to linear dynamics.
ClaimAudio · 24:57 · 3m
Philipp Herzig explains that while LLMs excel at unstructured text and images, enterprise planning requires predictions — demand forecasting, cash flow prediction, classification of customer payment behavior, regression of payment delays. LLMs are not designed for these tasks because they generate one token at a time in sequence-to-sequence modeling. Classical ML approaches like XGBoost work but don't scale because they require hiring data scientists and training separate models per country. SAP's RPT (Relational Pre-trained Transformer) applies the transformer architecture to structured tabular data, enabling accurate predictions with small amounts of data.
ClaimArticle · 22 words
The prompt should include an example of the tag format — like hierarchical product classifications such as 'Furniture / Living Room Furniture / Coffee Tables' — so the model's imagined tags look right.
ClaimArticle · 351 words
The teacher sets a quality ceiling, too large a capacity gap hurts transfer, a better base architecture can beat a larger one, and a 2025 Nature study showed behavioral traits can transfer through distillation even when data is filtered.
Mechanism◆Article · 135 words
The alternative to behavioral retrieval is semantic retrieval powered by embeddings, which place related items near one another in a high-dimensional space even without shared keywords. A dual-encoder design converts users and posts into points in the same space, letting platforms precompute post embeddings and run fast nearest-neighbor searches at request time.
Claim◆Audio · 16:08 · 2m
Eiso argues that model building is an engineering discipline, not theoretical magic. The Model Factory treats training as an end-to-end industrial process with streaming data, immutable data layers, and versioned code.