BEGIN:VCALENDAR
VERSION:2.0
PRODID:-//Sabre//Sabre VObject 4.5.8//EN
CALSCALE:GREGORIAN
LAST-MODIFIED:20241016T205237Z
BEGIN:VTIMEZONE
TZID:America/Los_Angeles
X-MICROSOFT-CDO-TZID:13
BEGIN:STANDARD
DTSTART:20231105T090000
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
END:STANDARD
BEGIN:STANDARD
DTSTART:20241103T090000
TZOFFSETFROM:-0700
TZOFFSETTO:-0800
TZNAME:PST
END:STANDARD
BEGIN:DAYLIGHT
DTSTART:20240310T100000
TZOFFSETFROM:-0800
TZOFFSETTO:-0700
TZNAME:PDT
END:DAYLIGHT
END:VTIMEZONE
BEGIN:VEVENT
UID:4dda4ee4-d1a8-4b8c-a7c5-2fc48519a013
DTSTAMP:20241016T205237Z
SUMMARY:Web API for Hybrid AI
DTSTART;TZID=America/Los_Angeles:20240925T100000
DTEND;TZID=America/Los_Angeles:20240925T110000
DESCRIPTION:https://www.w3.org/events/meetings/4dda4ee4-d1a8-4b8c-a7c5-2fc4
 8519a013/\n\nWeb apps are increasingly expected to gain access to a langua
 ge model. We are proposing Web APIs that allow web developers to directly 
 access both on-device and cloud-based language models\, and securely share
  user data between multiple apps when using these models.\n\nThe following
  are the APIs goals:\n\n- Provide web developers with a connection strateg
 y for accessing both on-device and cloud-based models. For example\, if no
  on-device models are available\, attempt to access cloud-based models. Co
 nversely\, if cloud-based models are unavailable\, try accessing on-device
  models.\n\n- Provide web developers with a storage strategy for sharing u
 ser's private data. For example\, one web app saves users' private data in
 to a local vector database. Another web app\, when accessing a on-device l
 anguage model\, can leverage this data through a local RAG system.\n\nThe 
 following are not within our scope of concern:\n\n- Design a uniform JavaS
 cript API for accessing browser-provided language models\, known as the [P
 rompt API](https://github.com/explainers-by-googlers/prompt-api)\, which i
 s currently being explored by Chrome's built-in AI team.\n\n- Issues faced
  by hybrid AI\, such as model management\, elasticity through hybrid AI\, 
 and user experience\, as this topic has already been discussed in [Hybrid 
 AI Presentations](https://github.com/webmachinelearning/hybrid-ai/tree/mai
 n/presentations) in the WebML IG\, and will be covered in the sessions on 
 [AI Model Management](https://github.com/w3c/tpac2024-breakouts/issues/15)
 .\n\nThis session will showcase how to use our proposed API for booking fl
 ights and hotels. It will also provide specific implementation details and
  references for these APIs. Example source code and implementation referen
 ces can be found on GitHub [web-hybrid-ai](https://github.com/kevinmoch/we
 b-hybrid-ai).\n\nAgenda\n\n**Chairs:**\nChunhui Mo\, Martin Alvarez-Espina
 r\n\n**Description:**\nWeb apps are increasingly expected to gain access t
 o a language model. We are proposing Web APIs that allow web developers to
  directly access both on-device and cloud-based language models\, and secu
 rely share user data between multiple apps when using these models.\n\nThe
  following are the APIs goals:\n\n- Provide web developers with a connecti
 on strategy for accessing both on-device and cloud-based models. For examp
 le\, if no on-device models are available\, attempt to access cloud-based 
 models. Conversely\, if cloud-based models are unavailable\, try accessing
  on-device models.\n\n- Provide web developers with a storage strategy for
  sharing user's private data. For example\, one web app saves users' priva
 te data into a local vector database. Another web app\, when accessing a o
 n-device language model\, can leverage this data through a local RAG syste
 m.\n\nThe following are not within our scope of concern:\n\n- Design a uni
 form JavaScript API for accessing browser-provided language models\, known
  as the [Prompt API](https://github.com/explainers-by-googlers/prompt-api)
 \, which is currently being explored by Chrome's built-in AI team.\n\n- Is
 sues faced by hybrid AI\, such as model management\, elasticity through hy
 brid AI\, and user experience\, as this topic has already been discussed i
 n [Hybrid AI Presentations](https://github.com/webmachinelearning/hybrid-a
 i/tree/main/presentations) in the WebML IG\, and will be covered in the se
 ssions on [AI Model Management](https://github.com/w3c/tpac2024-breakouts/
 issues/15).\n\nThis session will showcase how to use our proposed API for 
 booking flights and hotels. It will also provide specific implementation d
 etails and references for these APIs. Example source code and implementati
 on references can be found on GitHub [web-hybrid-ai](https://github.com/ke
 vinmoch/web-hybrid-ai).\n\n**Goal(s):**\nExplore the potential of our prop
 osed Web API for accessing hybrid AI through use case demonstrations and A
 PI implementations. Additionally\, discuss concrete steps for moving forwa
 rd.\n\n\n**Agenda:**\n1. Introduce the goals of the Web APIs we propose fo
 r hybrid AI (2m)\n2. Introduce Connection API (5m)\n3. Introduce Storage A
 PI (4m)\n4. A Showcase of Hybrid AI App (3m)\n5. Considerations for Connec
 tion Strategy (2m)\n6. Considerations for Storage Strategy (2m)\n7. Consid
 erations for Native OS APIs (2m)\n8. Discuss possible resolutions\, follow
 up actions and collaborations (20m)\n\n#### Links\n\n* [Web API for Hybrid
  AI - Slides](https://kevinmoch.github.io/web-hybrid-ai/)\n* [Web API for 
 Hybrid AI - Explainer](https://github.com/kevinmoch/web-hybrid-ai)\n\n**Ma
 terials:**\n- [slides](https://www.w3.org/2024/Talks/TPAC/breakouts/web-hy
 brid-ai/)\n- [minutes](https://www.w3.org/2024/09/25-web-hybrid-ai-minutes
 .html)\n- [Session proposal on GitHub](https://github.com/w3c/tpac2024-bre
 akouts/issues/44)\n\n**Track(s):**\n- AI
STATUS:CONFIRMED
CREATED:20240916T215330Z
LAST-MODIFIED:20241016T205237Z
SEQUENCE:1
ORGANIZER;CN=W3C Calendar;PARTSTAT=ACCEPTED;ROLE=NON-PARTICIPANT:mailto:nor
 eply@w3.org
LOCATION:2 Ballroom Level - California A
CATEGORIES:TPAC 2024,Breakout Sessions
END:VEVENT
END:VCALENDAR