slides · global inventory
Onion
Arc collection for the active corpus filters, with each row opening the slide detail view.
Matching slides 2,918
Current window 1,831-1,860
Pages 98
Rows per page 30
Arc
1,831-1,860 of 2,918
56
The slide highlights a specific AI research breakthrough and its practical application at Google.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
57
The slide describes a specific AI model (ATOMICA) and its application in drug discovery/biology.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
58
The slide highlights the technical architecture (MoLE) and the scale of training data (OMat24, OMat25, OMol25).2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
59
The slide highlights the shift from screening existing materials to generative design using diffusion models.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
60
The chart shows the evolution of chemical synthesis benchmark scores over time for various LLM providers.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
61
The slide highlights two specific research implementations: the Liverpool system and NC State's Rainbow platform.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
62
The slide describes a system that automates scientific method invention by treating it as a search problem.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
63
The slide compares Transformer++, Mamba, Hyena, and StripedHyena architectures in the context of genomic sequence modeling.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
64
The slide uses three distinct charts to demonstrate scaling laws, sequence space generation, and alignment benefits for protein LMs.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
65
The slide uses two charts to demonstrate that model success rates correlate strongly with similarity to training data.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
66
The slide presents performance metrics for AMIE vs PCPs using line charts showing accuracy over top-k.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
67
Slide 66 from the State of AI 2025 report.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)Loop · Zoom In1 hit
68
The slide highlights the shift from traditional autoregressive models to diffusion-based models for text generation.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
69
The slide explains a technical architecture (BLT) and compares its performance (BPB vs Training FLOPS) against standard BPE-based models.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
70
The slide uses a diagram to contrast 'No sink' vs 'Sink' behavior in token attention patterns.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
71
The slide highlights how LMArena data is being manipulated and how proprietary models have an unfair advantage in benchmark performance.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
72
The chart shows a negative correlation between capabilities score and attack failure rate, indicating that as models get more capable, they become more prone to failure in safety-critical tasks (WMDP).2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
73
Includes a screenshot of a ChatGPT conversation as a case study of sycophancy.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
74
The slide highlights the technical architecture (CNN, Transformer, Language Model) and the current limitations compared to invasive BCIs.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
75
The slide summarizes research findings on brain-model convergence, highlighting that early layers align with sensory cortices while later layers require more training to align with prefrontal regions.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
76
Includes specific cost estimates per hour for different neuroimaging modalities.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
77
The slide highlights a specific technical research development from Apple, detailing its architecture (4D sparse latent, pure-Transformer, 4D RoPE) and training scale.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
78
The slide discusses two specific technical approaches to robotic learning: neural rendering for 3D scene representation and next-token prediction for unified vision-language-action sequences.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
79
Discusses the trade-off between overfitting/compute costs and the ability to internalize complex physical dynamics.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
80
Discusses Molmo-Act (AI2), Gemini Robotics 1.5 (GDM), and MIT's 'Teaching LLMs to Plan' research.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
81
The slide uses a diagram to explain the EMMA model architecture, showing inputs (text/vision) feeding into the model and outputs (reasoning/planning).2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
82
The slide includes a detailed performance table comparing various agents across 15 games.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
83
The slide presents a technical argument for using smaller models (1-9B parameters) in agentic workflows to reduce costs and latency, while maintaining an 'escape hatch' to larger models.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
84
The slide uses a specific case study of Sakana AI to illustrate the broader point about the need for domain-expert audits in AI research.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit
85
The slide uses a metaphor (USB-C) to explain the standardization of AI tool integration.2025 Air Street Capital The State of AI Report 2025 · Air Street Capital
Consulting deck VC research Beat · The Facts (What)1 hit