Lifeboat Foundation: Safeguarding Humanity
Toggle light / dark theme

Blog

Category: robotics/AI – Page 74

Token Caching Secrets: Cut AI Enterprise Costs By 80%

Token caching can cut enterprise AI costs by 80%

In 2026, the biggest cost advantage isn’t choosing a cheaper model — it’s token caching.

When the same system prompts or knowledge base get reused, providers discount that input by 80–90%. Claude Sonnet 5, for example, drops from $2.00 to $0.20 per million tokens on cached content.

Enterprises with heavy RAG or multi-turn workflows are leaving $400K–$520K on the table annually by not using it.

Key takeaway for procurement:

Stop comparing raw token prices. Start modeling cost-per-result on your actual query patterns. Smaller models with strong caching often win.

Full analysis:

Continue reading “Token Caching Secrets: Cut AI Enterprise Costs By 80%” | >

Why World Models Could Change Robotics, 3D, and Creativity

World Labs co-founders Fei-Fei Li, Justin Johnson, and Ben Mildenhall join a16z General Partner Martin Casado to discuss Atlas, their latest world model, and what it reveals about the pursuit of spatial intelligence. At the center of Atlas is what the team calls “new view prediction”: given images or views of a scene, the model predicts what that environment should look like from a different position in space and time. This brings generation and 3D reconstruction into the same model, and raises a broader question about whether predicting views could become a useful primitive for understanding the physical world. They discuss the technical bets behind the model, what it can and can’t yet capture, and the importance of dynamics, editability, and simulation as world models develop. The conversation also explores applications in creative work, architecture, and robotics, where Fei-Fei argues that one of today’s biggest constraints is access to real-world training data. Timestamps: 00:00 — Intro 00:51 — What Atlas Is & Why It Matters 05:15 — Is This a Scaled-Up Video Model or a New Architecture? 08:15 — Spatial Intelligence & Why New View Prediction Matters 21:27 — Did You Know It Was Going to Work? 24:42 — Use Cases: Creatives, Games & Robotics 35:21 — The Elephant in the Room: Video Models vs World Models 37:55 — Will We Get 4D Video You Can Walk Around In? 42:22 — Why New View Prediction Is the Next Token Prediction Resources: Follow Fei-Fei Li on X: https://twitter.com/drfeifei Follow Justin Johnson on X: https://twitter.com/jcjohnss Follow Ben Mildenhall on X: https://twitter.com/BenMildenhall Follow Martin Casado on X: https://twitter.com/martin_casado Learn more about Atlas: https://www.worldlabs.ai/blog/atlas Stay Updated: If you enjoyed this episode, be sure to like, subscribe, and share with your friends! Find a16z on X: / a16z Find a16z on LinkedIn: / a16z Listen to the a16z Show on Spotify: https://open.spotify.com/show/5bC65RD… to the a16z Show on Apple Podcasts: https://podcasts.apple.com/us/podcast… Follow our host: https://twitter.com/eriktorenberg Please note that the content here is for informational purposes only; should NOT be taken as legal, business, tax, or investment advice or be used to evaluate any investment or security; and is not directed at any investors or potential investors in any a16z fund. a16z and its affiliates may maintain investments in the companies discussed. For more details please see http://a16z.com/disclosures.

New AI tool maps the hidden universe of small molecules

The human body and its gut microbiome produce thousands of small molecules that shape how the body functions—influencing immunity, metabolism and more. Identifying what those molecules actually are has been one of the biomedical sciences’ most persistent bottlenecks. More than 80% of compounds detected in a typical biological sample cannot be matched to any known structure using current methods.

Researchers at the Boyce Thompson Institute (BTI) and Cornell University have developed a tool that begins to change that. AIMe, short for AI Molecule Explorer, uses a form of artificial intelligence called neuro-symbolic AI to predict, organize and search the mass spectra of more than 100 million known small organic molecules—effectively building a vast searchable map of chemical space that can accelerate hypothesis generation and compound identification.

The work is a collaboration between Frank Schroeder, professor at BTI and in Cornell’s Department of Chemistry and Chemical Biology, and Carla Gomes, professor of computing and information science and director of Cornell’s AI for Science Institute.

AI digital twins struggle to predict human behavior, creating ‘funhouse mirror’ distortions

While many fear artificial intelligence will replace humans, using AI to take over some human roles has benefits. Companies can use the technology to conduct surveys and polls, while behavioral scientists can run experiments on digital twins to gather faster insights without risking harm or distress to real participants.

But there’s a huge snag.

Today’s digital twins are just not up to it. New research published in the journal Science Advances found that digital twins seriously misrepresent human behavior.

The composer still making music four years after his death — thanks to an artificial brain

I love this kind of synergy between art and technology — another demonstration of how these two concepts really aren’t separate after all. People build up so many experiences over their lives, each moment layering on the previous one to create a uniquely complex tapestry. When we die, all of this is lost to thermal noise. While this particular artistic iteration may not exactly capture memory engrams, I hope that future posthumous neuro-arts find a way to transmit more of the original person’s experiential tapestry. I think it is a lovely idea to try to keep at least part of a person’s uniqueness growing, changing, and creating even after death.

In Australia, a team of artists and scientists have resurrected the US composer Alvin Lucier. It raises a storm of questions about AI and authorship – and it’s also incredibly beautiful.

The statement means that the $12.93 billion acquisition price is a piece of technical inside humor

It literally encodes the “Hugging Face” emoji (🤗) and a nod to NVIDIA’s brand color directly into the financial figure, serving as a geeky tribute to the community that made Hugging Face strategically valuable.

What is the Easter egg? The $12,930,300,000 price announced for Nvidia’s acquisition of Hugging Face contains two references to the companies. Hugging Face co-founder Thomas Wolf invited readers to find both meanings in the number, turning an otherwise unusually precise acquisition price into a small piece of technical and visual wordplay.

The Hugging Face reference: The sequence 129,303 is the decimal representation of Unicode code point U+1F917, the 🤗 emoji officially named “Hugging Face”. The number therefore encodes the company’s distinctive name and icon directly into the announced consideration. It is an especially fitting reference for a company whose brand is built around the emoji and whose platform has become a central meeting place for the open machine-learning community.

The Nvidia reference: The same number also works as a hexadecimal colour code: #129303. That produces a vivid green close to Nvidia’s familiar brand colour. Earlier online theories split the digits into 12, 93 and 3, suggesting references to Nvidia’s $12 IPO price, its 1993 founding and its three founders; however, Clément Delangue pointed readers towards the colour-code explanation. He also said the detail showed that the price was never the most important thing for either side.

Japan Is Launching a Probe to Collect the FirstEver Samples From a Martian Moon

Once on the surface, a robotic arm will drill cylindrical tubes into the surface to collect samples. The probe is also equipped with a device first developed by NASA that uses a stream of nitrogen gas to collect fine particles from the very top layer of dust. Whether or not Phobos originated from an impact with the planet, experts believe Martian sand can be obtained from the lunar surface. Over time, meteorite impacts have blasted materials off Mars, coating its moons. Scientists expect about 0.1 percent of the samples collected will be debris from Mars itself.

Costing about $345 million US, this mission marks Japan’s first Mars probe launch in 28 years. Beyond the scientific insights it’s expected to provide, from an engineering perspective, the mission aims to test new technologies that could aid future sampling and round-trip travel to Mars.

MMX is scheduled to launch from the Tanegashima Space Center on October 20, 2026.