Rezultati pretraživanja
  1. 1. velj
    Prikaži ovu nit
  2. 28. sij
  3. 26. ožu 2019.

    Video from the AI Colloquium is now up on our YouTube channel, DARPAtv. The panel on AI was particularly interesting: . Thx to , , 's Richard Danzig & Patrick Lin from Ethics + Emerging Sciences Group .

  4. 19. ožu 2019.

    Today, / launched to showcase government-wide accomplishments in , including strategic documents, fact sheets, programs, and more from throughout the Federal R&D enterprise, including DARPA's AI Next campaign!

  5. 12. ožu 2019.

    . researchers supporting our Lifelong Learning Machines prgm created an AI-controlled robotic limb driven by animal-like tendons & capable of teaching itself a walking task. Research is featured on cover of Nature Machine Intelligence.

    Prikaži ovu nit
  6. 11. ožu 2019.

    In an interview with , DARPA's Valerie Browning explained the investment strategy behind DARPA's $2 billion campaign and how the agency aims to transform computers into trusted, collaborative partners.

  7. 8. ožu 2019.

    With a coordinated strategy for AI, progress on enabling tech that makes AI safer, more robust, more adaptable, and more intuitive, plus novel applications in areas like chemistry, manufacturing, and scientific discovery, the U.S. is on track to lead the 3rd Wave of AI.

    Prikaži ovu nit
  8. 8. ožu 2019.

    One of the underlying themes of our AI Colloquium was delivering the capabilities needed to turn J.C.R. Licklider's vision for human-machine teaming into reality. 59 years later, it remains a work in progress, but the research currently underway is bringing us closer.

    Prikaži ovu nit
  9. 7. ožu 2019.

    None of the progress made in AI was brought about by DARPA alone. There have been very productive public-private partnerships, sustained funding from key players & fresh leadership from & to coordinate resources in support of a national AI strategy.

    Prikaži ovu nit
  10. 7. ožu 2019.

    Right now, at least 1/3 of DARPA programs involve AI in some fashion. We're bringing about 3rd Wave of AI — which we THINK is contextual reasoning — but check back in 20 yrs to see how it actually manifests. In any case, AI is clearly relevant to mission.

    Prikaži ovu nit
  11. 7. ožu 2019.

    However, there's much more work left to do. The field of AI has been marked by successive waves of hype, followed by "AI winters" when expectations weren't met & funding declined. DARPA & some other funders made sustained investment to advance the foundations of AI.

    Prikaži ovu nit
  12. 7. ožu 2019.

    DARPA Deputy Director: After 50 years of pushing & striving, the field of AI became an overnight success in the past decade. DARPA’s investments in AI began even before 1st episode, and since then have involved tens, if not hundreds of thousands of engineers.

    Prikaži ovu nit
  13. 7. ožu 2019.

    DARPA's Media Forensics program is developing scalable, automated tools to spot fakes, looking at digital, physical & semantic integrity of images & video. In the future, it may become possible to rapidly flag & filter manipulated content.

    Prikaži ovu nit
  14. 7. ožu 2019.

    In every conference, there's going to be one session no one cares about... It's time to discuss media manipulation & how AI can help identify fakes. Sorry. Left: earliest known photo manipulation from 1865. "Honest Abe" isn't so honest here. Right: AI-generated faces.

    Prikaži ovu nit
  15. 7. ožu 2019.

    Trust is essential to adoption of AI-enabled tech, but current assurance approaches for cyber-physical systems assume systems don’t change once deployed. With arrival of learning-enabled systems, safety assurance needs to move beyond design time into operation time.

    Prikaži ovu nit
  16. 7. ožu 2019.

    Engineers routinely use high-fidelity simulations to design for tech such as jets, cars & chips. However, robust design is elusive in domains like synbio, protein design & materials chem due to lack of high-fidelity models. AI can help overcome lack of complete models.

    Prikaži ovu nit
  17. 7. ožu 2019.

    You cannot prepare for every eventuality when creating machine learning training sets, yet ML systems are essentially frozen in time once they’re trained & fielded. This means they can malfunction in unanticipated circumstances.

    Prikaži ovu nit
  18. 7. ožu 2019.

    Behold the of AI! DARPA is working to embed physics & prior knowledge in AI models to constrain problem spaces & reduce need for labeled training data.

    Prikaži ovu nit
  19. 7. ožu 2019.

    Revolutionizing Prosthetics is leading charge in fusion of AI/BCI. Researchers explore how to generalize control & reason across task primitives — steps that go into task completion (e.g., reach, grasp, lift). Primitives can be recombined to facilitate new tasks.

    Prikaži ovu nit
  20. 7. ožu 2019.

    Addition of AI to BCI is helpful to reduce amount of training & info required to make a BCI facilitate tasks. AI can help through features like sensors, perception & trajectory planning, allowing a user to complete a task w/ far less concentration & training.

    Prikaži ovu nit

Čini se da učitavanje traje već neko vrijeme.

Twitter je možda preopterećen ili ima kratkotrajnih poteškoća u radu. Pokušajte ponovno ili potražite dodatne informacije u odjeljku Status Twittera.