When the Tribune Parameter Consortium on July 28-31, San Jose will have a 2025 All -Hands Conference in San Jose and Exhibition, TPC 25,, Competitive Scientists, System Architects, Policy Leaders, and Research AI’s stakeholders, and a question for the government.
For more than three days, participants of TPC 25 will dive into the building’s agents and hands -related lessons on Hikathon. Planery speakers like Argon’s Rick Stevens and Rican’s Satoshi Matsuka will outline the priorities of scientific AI, while 30 30 breakout sessions of cancer research will be from LLM to supercompointing energy efficiency. As a result, the Part Workshop, the Part Strategy Summit, and the preliminary look at how open, Frontier -scale can renew AI science in the next decade.
Agent tutorials and hackathone
Organizers have developed the program around three straightforward goals: expand the partnership community through tutorials and hackets, present a broad theory in a series of full interaction, and preview and make projects during the breakout sessions slate.
For a day and a half, the interests of making Frontier AI more accessible for researchers will have lessons and hecatches to bring more hands on the deck. Parallel to the lesson, Hikathon will connect newcomers to all three teams with experienced developers. Each group will create a simple agent inside an agent framework, such as a DOI folk -up boot or lightweight web search help. This setup will ensure a healthy mixture of skill levels, which organizers hope to give rise to learning.
Focus on agents reflects on how new AWS bedrock agent core, Microsoft Discovery, deep research of Openi, and Google Deep Mind’s AI co -scientists are becoming standard in science -work workfloose science through a platform. By the end of the session, participants should be discharged with a practical experience they can apply for their research.
Full Highlights: Leader Make a Map of the Road
The full session will present a major view of the priorities of the World Research Computing Community. The Director of the Associate Laboratory in Argon and an Avary person, Rick Stevens, will open a full series with “Inventing Discovery: Artificial Super – Intelligence Accelerating Science”. Stevens will discuss how Frontier – Scale Model Task – has moved beyond the present to help create existing scientific questions and methods as specific AI tools. In his inaugural conversation, he will find out how quickly the benefits of speculation, imitation, and the explanation of statistics are rapidly confronted by slow measures such as chemistry, fabric and fieldwork. Baying from the law of Amidahl, he will outline the challenge that instead of the same jump in calculation, the entire research workflow will be outlined as a restoration of the flu. Stevens will also deal with practical questions: How to set goals that reflect human preferences, keep the results reliable when artificial data is used, and decide which tasks still need human skills. He plans to close the society with concrete steps to navigate the new landscape simultaneously.
AI researchers such as Rick Stevens, Satoshi Mitsuka, and Charlie Catellet will participate in TPC 25. (Source: TPC 25)
The AWS’s Thirin will make the Pelgino Hybrid HPC a matter, or – to connect the Premies system to the cloud scope, as a practical path, connect the Premies System. His conversation will show how AI tools, classical imitation, and early quantum resources can share a workflow, which can produce short examples from weather forecasts, computational fluid dynamics, and molecular modeling, which has promoted computing jobs without excessive costs.
The Rican Center for Competition Science will present the outline of Satoshi Mitsuka, who still stands between large language models and everyday scientific works. Rican’s AI4S Science Program and recent runs on Fogaku and Frontier Super Computers, they will discuss multi -tarbite data handling, making nerve surrogates with physics solors, and spreading workflows in CPU, GPU, and early quantum hardware. The preliminary results of climate modeling, 3-D-imaging, and material discovery will show up to 6.9x speed-ups, and Matsuka will shut down open questions around the Provision and Human Surveillance for Patastical and Bound ML Stics.
Other planers will examine the technical and moral concerns as AI moves deeply in science. The Ian Foster of Argon will outline an “AI – Local Discovery Platform” that will marry the trillion – parameter model, marry the graph of knowledge and independent labs. Richardo Beaza of the Barcelona Superping Center will present an irresponsible AI survey in the form of prejudice and notoriety before maping of governance and regulatory tools. Microsoft’s Pratt will ask Chengpa when AI agents are made everywhere, who is the agency, and what does that mean to control ownership and results. Kartak Dorissimi (University of Michigan) will outline a proposed “active in conference” architecture, which is a layered system that constantly connects a model of counter -arguing with real -world verification, to shut down three stubborn distributions in research: summary space, reasoning space, and reality space.
Together, such a complete session and many others have planned for the TPC, helping participants to assess the successes and challenges of Frontier -Scale AI in the scientific process.
Breakout tracks: BOF sessions and turning Ideas into working groups in power negotiations
(Source: Shutter stock)
The third target of TPC 25 is to advance cooperation focused through the afternoon breakout sessions. Six rooms and five -time blocks spread 30 breakout slots. Most sessions working – group form will follow: 90 – Minute blocks that open with four to six power negotiations and leave time for concentrated conversation. Some tracks are spread over numerous blocks, giving a follow -up session room to improve ideas.
Birds – a – a – feider mats will run in parallel. The topic tracks covers both the research requirements and long -term infrastructure questions. The domain science life is deep in the sciences lane, where three sessions review the AI for cancer, agent bioral workflow, and the foundation model for biology. Heavenly power negotiations will discuss topics like peptide – binder design, genome -scale language model, and precision – health pipelines. Other sessions will deal with energy efficiency for AI on supercumping sites and plan for the Foundation model that represents the electric – grid distribution system. Organizers expect that at the end of the meeting, several of these informal meetings will be fully fully in full TPC working groups.
Why TPC25 AI -Significant Science
With practical sessions, full conversations and breakouts that can convert ideas into prototypes, TPC will provide a clear look at Frontier AI and its future guidance for science. Lessons and hackers will help participants gain new skills, while full conversation will outline AI’s influence on research. Birds of winging meetings and power negotiations will move forward. Researchers, system builders, and anyone who follows the effects of AI on science, should not be remembered by the San Jose gathering.
To see the agenda and learn more about the trillian parameter consortium and TPC 25, see https://tpc25.org And register https://tpc25.org/register.
Relevant
