10:18:12 <RRSAgent> RRSAgent has joined #ai-agents
10:18:16 <RRSAgent> logging to https://www.w3.org/2025/03/26-ai-agents-irc
10:18:16 <breakout-bot> RRSAgent, do not leave
10:18:17 <breakout-bot> RRSAgent, this meeting spans midnight
10:18:17 <breakout-bot> RRSAgent, make logs public
10:18:18 <breakout-bot> Meeting: How would AI Agents change the Web platform?
10:18:18 <breakout-bot> Chair: Dominique Hazaël-Massieux
10:18:18 <breakout-bot> Agenda: https://github.com/w3c/breakouts-day-2025/issues/7
10:18:18 <Zakim> Zakim has joined #ai-agents
10:18:19 <breakout-bot> Zakim, clear agenda
10:18:19 <Zakim> agenda cleared
10:18:19 <breakout-bot> Zakim, agenda+ Pick a scribe
10:18:21 <Zakim> agendum 1 added
10:18:21 <breakout-bot> Zakim, agenda+ Reminders: code of conduct, health policies, recorded session policy
10:18:21 <Zakim> agendum 2 added
10:18:21 <breakout-bot> Zakim, agenda+ Goal of this session
10:18:22 <Zakim> agendum 3 added
10:18:22 <breakout-bot> Zakim, agenda+ Discussion
10:18:22 <Zakim> agendum 4 added
10:18:22 <breakout-bot> Zakim, agenda+ Next steps / where discussion continues
10:18:24 <Zakim> agendum 5 added
10:18:25 <breakout-bot> Zakim, agenda+ Adjourn / Use IRC command: Zakim, end meeting
10:18:25 <Zakim> agendum 6 added
10:18:25 <breakout-bot> breakout-bot has left #ai-agents
10:26:20 <tidoust> tidoust has joined #ai-agents
13:21:04 <tantek> tantek has joined #ai-agents
13:40:13 <ktoumura> ktoumura has joined #ai-agents
13:53:10 <ktoumura> ktoumura has joined #ai-agents
13:53:10 <tantek> tantek has joined #ai-agents
13:59:24 <bkardell_> bkardell_ has joined #ai-agents
14:00:02 <bkardell_> hi!
14:00:04 <dom> Present+ Brian, Anssi, Hadley, DavidBaron, PLH, Dom
14:00:56 <dbaron> dbaron has joined #ai-agents
14:01:03 <danbri> danbri has joined #ai-agents
14:01:15 <hadleybeeman> hadleybeeman has joined #ai-agents
14:01:15 <denis> denis has joined #ai-agents
14:01:15 <plh> plh has joined #ai-agents
14:01:20 <plh> present+
14:01:37 <plh> zakim, who is here?
14:01:37 <Zakim> Present: Brian, Anssi, Hadley, DavidBaron, PLH, Dom
14:01:38 <Zakim> On IRC I see plh, denis, hadleybeeman, danbri, dbaron, bkardell_, tantek, ktoumura, tidoust, Zakim, RRSAgent, dom
14:02:04 <hyojin> hyojin has joined #ai-agents
14:02:21 <plh> present+ hadleybeeman , dbaron , ivan_herman
14:02:24 <anssik> anssik has joined #ai-agents
14:02:36 <plh> present+ Kunihiko_Toumura
14:02:37 <hadleybeeman> Present+ DanBri
14:03:03 <anssik> Present+ Anssi_Kostiainen
14:03:05 <hyojin> Present+ Hyojin Song
14:03:10 <Mizushima> Mizushima has joined #ai-agents
14:03:23 <plh> present+ Larry_Zhao
14:03:43 <denis> present+
14:03:43 <plh> present+ Iris_ben
14:03:47 <plh> present+ Scott_Hendrickson
14:03:53 <plh> present+ Ari
14:04:07 <plh> present+ Kaustubha
14:04:08 <plh> scribe+
14:04:16 <McCool> McCool has joined #ai-agents
14:04:32 <Mizushima> present+ Tomoaki_Mizushima
14:04:33 <McCool> passcode in invite seems to be incorrect for zoom
14:04:44 <ivan> ivan has joined #ai-agents
14:05:18 <plh> -> https://www.w3.org/2025/Talks/dhm-ai-agents/ Slides
14:05:27 <plh> [slide 2]
14:05:42 <plh> Dom: AI is in the news noawadays
14:05:48 <RobSmith> RobSmith has joined #ai-agents
14:05:51 <plh> ... agentic AI, etc.
14:05:54 <McCool> passcode in invite is incorrect
14:06:22 <plh> passcode 075582
14:06:41 <McCool> thx
14:06:47 <plh> [slide 3]
14:07:08 <plh> dom: let's a look at how LLMs interact on the Web
14:07:10 <plh> [slide 4]
14:07:32 <plh> dom: chatbot is simple and classic
14:07:36 <plh> [slide 5]
14:07:50 <plh> dom: evolution: LLM on the client side
14:08:04 <plh> ... the web site download an LLM in your browser
14:08:11 <plh> ... (WebGPU, WebNN, WASM)
14:08:25 <plh> [slide 6]
14:08:56 <plh> dom: variation: Task APIs. You interact with the embedded LLM, which you access through well scoped APIs
14:09:16 <plh> ... the LLM is operating on the user device and linked to specific tasks
14:09:20 <plh> [slide 7]
14:09:38 <plh> dom: you delegate to the LLM
14:10:00 <plh> ... instead of searching for a shop to buy shoes
14:10:16 <plh> ... you ask the LLM to do it for you
14:10:47 <plh> ... so analizing search results, initiating the purchase, etc.
14:11:20 <plh> ... so the browser is no longer the core center of  user interaction
14:11:30 <plh> ... the LLM uses the browser
14:11:36 <plh> [slide 8]
14:12:02 <plh> dom: the LLM could on server-side for those cases
14:12:07 <plh> s/coul/could be/
14:12:30 <plh> ... so those were appromixation of LLM variations
14:12:36 <plh> [slide 9]
14:12:51 <plh> dom: home assistants have been deployed for sometime
14:13:06 <plh> ... those had some charateristics of AI agents
14:13:21 <plh> ... some are shipping chatbots as part of the UI
14:13:33 <plh> ... interacting with the webpage as part of the browser
14:13:33 <danbri> (minor nitpic) re MCP, it's not specifically an http API, rather they initially define only stdio and HTTP transports https://modelcontextprotocol.io/docs/concepts/architecture ... initially same local host, with a view to extending to networked services later.  It's basically some JSON-RPC interfaces. I've been experimenting with a WebRTC data
14:13:33 <danbri> channel binding.
14:13:54 <plh> ... recent announcement from firefox having an LLM in a browser extension
14:14:33 <plh> ... [deep research agents]
14:15:04 <plh> ... uses AI agent to test web content
14:15:17 <McCool> q+
14:15:41 <plh> ack mc
14:16:04 <plh> McCool: an other option on architecture: using a nearby computer
14:16:19 <plh> ... has some privacy/security advantages
14:16:49 <handy9> handy9 has joined #ai-agents
14:16:50 <plh> ... giving raw access to AI agents to your browsing has privacy implications
14:17:13 <plh> [slide 10]
14:17:39 <plh> dom: create additional intermediate between the user and the web
14:17:58 <hadleybeeman> We the TAG have been talking a lot about when this becomes a user agent.
14:18:17 <plh> ... switching from clicking to access services to a descriptive interaction model
14:18:42 <plh> ... non-deterministic: output will differ
14:19:29 <plh> ... boundaries between robots and user agents. robots are fully automated, while a browser is controlled by a user
14:19:55 <plh> ... AI agent is not quite a crawler but not exactly a user agent either
14:20:02 <McCool> two main points: where the LLM runs, and what it has access to (which has two aspects: what data it can use, and what it can do - delegated rights, etc)
14:20:05 <plh> ... should those agents respect robots.txt ?
14:20:16 <plh> [slide 11]
14:20:30 <plh> dom: impact on web platform
14:20:44 <plh> ... MCP to interact with online services
14:21:09 <plh> ... it's an emerging specification
14:21:39 <ivan> present+
14:21:40 <plh> ... a new of website are looking at providing LLM-tailored content
14:22:06 <plh> ... seems similar to early discussions on mobile and a11y
14:22:23 <plh> ... efforts looking at sitemaps for AI agents
14:22:25 <danbri> LLMs.txt spec - https://llmstxt.org/
14:22:36 <plh> ... content gets customized for LLMs
14:23:16 <plh> ... LLM and CATPCHA: LLMs may prevented from solving those
14:23:20 <cpn> cpn has joined #ai-agents
14:23:25 <plh> [slide 12]
14:23:34 <plh> dom: risks
14:24:06 <plh> ... security: hallucinations: LLM may be doing something as completely different than expected. the more access, the more risk...
14:24:48 <plh> ... no strong difference between what they're supposed to do as a program vs on behalf of the user
14:25:09 <plh> ... eg sending your credentials to third party
14:25:20 <plh> ... privacy
14:25:40 <plh> ... with distributed architecture, it may make privacy more complex
14:25:59 <plh> ... ecosystem
14:26:16 <plh> ... diluted the end user intent from the content provider
14:26:38 <plh> ... content is monetized based on the assumption that ads will be seen by users
14:26:51 <plh> .... if you reduce it, it will change the dynamics
14:27:02 <plh> [slide 13]
14:27:16 <plh> dom: what did I miss?
14:27:28 <plh> ... input on the role of W3C in this space?
14:27:37 <plh> ... how to organize a discussion
14:28:01 <Dingwei> Dingwei has joined #ai-agents
14:28:03 <danbri> q+ to mention multimodal models (beyond LLM text only) and webrtc
14:28:03 <plh> ... should we wait before it's a fast moving space?
14:28:20 <plh> q+ to talk about MCP
14:28:49 <plh> ... if you have contacts for us, we're interested
14:29:00 <plh> [end of presentation]
14:29:03 <dom> q?
14:29:04 <plh> q?
14:29:06 <anssik> q+
14:29:10 <plh> ack dan
14:29:10 <Zakim> danbri, you wanted to mention multimodal models (beyond LLM text only) and webrtc
14:29:21 <plh> danbri: nice to see you're on top of all of this.
14:30:05 <zolkis> zolkis has joined #ai-agents
14:30:11 <McCool> q+
14:30:15 <zolkis> q+
14:30:19 <plh> ... it's good to focus on multimodal. some of those can understand images, sounds, etc. streaming voice system from google, etc. they're starting to use webrtc
14:30:26 <plh> ... using raw sockets
14:30:47 <plh> ... my advise would be to build on top of those uses
14:31:08 <plh> ... rather rather the general AI
14:31:17 <McCool> (use cases would be helpful - e.g. accessibility, describe images; language transcription; PNG->SVG conversion; etc; BUT evolving quickly; also, not just LLMs)
14:31:19 <plh> dom: taking into account greater diversity of needs
14:31:20 <danbri> webrtc openai native support, https://platform.openai.com/docs/guides/realtime - webrtc for realtime api (also websockets);
14:31:20 <danbri> 3p gemini integration https://github.com/pipecat-ai/gemini-webrtc-web-simple https://ai.google.dev/gemini-api/docs/live
14:31:22 <dom> scribe+
14:31:49 <dom> PLH: my understanding of MCP is that it is not meant to replace HTTP, but is an RCP protocol to be used locally
14:32:09 <dom> ... it comes with no security layer - e.g. it reuses the user credentials
14:32:47 <plh> dom: where do you put the boundaries?
14:32:50 <McCool> q?
14:32:55 <plh> ... if browsers have an LLM
14:32:56 <dom> scribe-
14:33:01 <dom> ack plh
14:33:01 <Zakim> plh, you wanted to talk about MCP
14:33:05 <dom> ack anssik
14:33:19 <tomayac> tomayac has joined #ai-agents
14:33:25 <plh> anssi: the potential to disrupt the current moneytization of the web
14:33:27 <tomayac> present+
14:33:39 <plh> ... "nobydo knows you're an AI agent"
14:33:59 <plh> ... but a lot of positive here, useful use cases.
14:34:21 <plh> ... it's still valuable for the community to keep up with this development
14:34:35 <plh> .... standardize pieces maybe
14:34:49 <plh> ... w3c cares about horizontals
14:35:01 <plh> ... ethical being one of them
14:35:09 <danbri> on plh's MCP summary, it captures where MCP is today but there's a roadmap https://modelcontextprotocol.io/development/roadmap and non-local non-public services are clearly on their way.
14:35:09 <danbri> """Our top priority is improving remote MCP connections, allowing clients to securely connect to MCP servers over the internet. Key initiatives include:
14:35:09 <danbri> Authentication & Authorization: Adding standardized auth capabilities, particularly focused on OAuth 2.0 support.
14:35:09 <danbri> Service Discovery: Defining how clients can discover and connect to remote MCP servers.
14:35:10 <anssik> https://www.gov.uk/government/publications/international-ai-safety-report-2025
14:35:11 <danbri> Stateless Operations: Thinking about whether MCP could encompass serverless environments too, where they will need to be mostly stateless."""
14:35:36 <plh> McCool: looking at use cases and gaps into our specifications
14:35:59 <Harry4> Harry4 has joined #ai-agents
14:36:09 <plh> .... writing use cases and thinking their impact on use cases would be useful. not sure who should do it.
14:36:20 <plh> ... lots of positive around a11y
14:36:29 <hadleybeeman> q+
14:36:38 <plh> dom: collecting use cases with a cg?
14:36:46 <plh> ... I did not include a11y in my slides
14:36:55 <plh> ... but agree it's an important aspect
14:36:56 <dom> ack McCool
14:37:09 <plh> McCool: image description, reorganizing layouts, etc.
14:37:18 <plh> ... but it's opended and it is evolving
14:37:28 <dom> ack zolkis
14:37:46 <plh> zolkis: user agents vs AI agents, AI workflows, normal workflows
14:38:14 <plh> ... in a web browser we need to think about process architecture changes vs thinking about workflows
14:38:28 <plh> ... not clear where AI agents fall into those spaces
14:38:54 <plh> dom: good input to look into. I welcome pointers
14:39:05 <plh> hadleybeeman: [muted]
14:39:12 <dom> RRSAgent, draft minutes
14:39:14 <RRSAgent> I have made the request to generate https://www.w3.org/2025/03/26-ai-agents-minutes.html dom
14:39:21 <plh> ... [still muted]
14:39:30 <plh> ... [lots of good things that we can't hear]
14:39:43 <shendrickson> shendrickson has joined #ai-agents
14:40:19 <plh> .... thanks for the work done so far
14:40:31 <plh> ... TAG has AI in its radar
14:40:41 <plh> ... and what can we say and do
14:40:52 <plh> ... we don't have consensus yet on a statement but working towards it
14:41:28 <plh> ... I'm focusing otm: LLMs as user agents
14:41:55 <plh> .... they should have the same responsibilities than user agents
14:41:55 <dom> [related to User Agents duties & TAG: https://w3ctag.github.io/user-agents/ ]
14:41:59 <plh> ... going back to ethical web principles and web design principles
14:42:21 <plh> ... making sure those can adapt to new technologies
14:42:29 <zolkis> A good primer/summary (further material on the channel) on AI workflows vs AI agents (and you can ask Gemini/Perplexity/ChatGPT as well): https://www.youtube.com/watch?v=tx5OapbK-8A
14:42:43 <plh> ... wanted to hear what I'm missing and what else others are doing
14:42:49 <McCool> q+
14:43:03 <plh> dom: where should we continue the conversation?
14:43:10 <plh> McCool: should we have a workshop?
14:43:16 <anssik> q+
14:43:34 <plh> dom: i've been struggling to find the right way to move forward, rather than just watching
14:43:43 <plh> ... re workshop: not clear to me that we know enough to have a productive conversation
14:43:49 <shendrickson> q+
14:43:56 <plh> McCool: one breakout isn't enough
14:44:05 <plh> .... workshop, cg, tpac
14:44:14 <dom> ack hadleybeeman
14:44:16 <dom> ack McCool
14:44:25 <dom> ack anssik
14:44:30 <plh> anssik: I can talk to the web machine learning wg/cg
14:44:35 <plh> .... to see if there is interest
14:44:48 <plh> ... also we can use time in the cg if you would like
14:44:54 <hadleybeeman> I should’ve mentioned the Web ML group — also written wonderful docs. I got a lot out of the use cases.
14:45:07 <plh> ... interested in hearing positions from the browser vendors or other shipping products
14:45:22 <plh> q+ to mention AC meeting
14:45:29 <danbri> q+ to suggest a WG-centric workflow for a W3C response to the AI situation
14:46:24 <plh> shendrickson: on progressing this work: IETF looking into authenticating agents to web sites
14:46:39 <plh> .... we should try to find a few more of those
14:46:46 <plh> ... but space is very broad
14:47:01 <plh> .... we should try to focus on use cases around agents
14:47:25 <plh> dom: intersection is indeed difficult
14:47:25 <dom> ack shendrickson
14:47:27 <dom> ack plh
14:47:27 <Zakim> plh, you wanted to mention AC meeting
14:47:33 <dom> scribe+
14:47:49 <dom> PLH:  I want an AI agent to scribe in my stead
14:47:59 <dom> ... there will be another AI-related session at the upcoming AC meeting
14:48:17 <dom> scribe-
14:48:20 <dom> ack danbri
14:48:20 <Zakim> danbri, you wanted to suggest a WG-centric workflow for a W3C response to the AI situation
14:48:37 <plh> danbri: this trend is as big as when the web started. all groups should think about it
14:48:51 <plh> ... I would encourage to have a top-level to grassroot approach
14:48:52 <RobSmith> q+ to mention parallels with search engines
14:49:13 <plh> ... asking every group for a one-page on how AI impacts them
14:49:34 <plh> dom: I love the suggestion. it grounds to our current work
14:49:56 <plh> ... for AI in general or for AI agent evolution?
14:50:09 <plh> danbri: for everything
14:50:21 <plh> ... is there anyone thinking about forms?
14:50:41 <plh> dom: this might also serve as a good ground before a workshop
14:51:07 <plh> RobSmith: parallels with existing systems: sitemaps, etc.
14:51:21 <plh> ... search engines are another one
14:51:29 <plh> ... SEOs optimization
14:51:39 <plh> ... taking some parallel approachs for AI
14:51:57 <plh> ... AI is an extension of robots
14:52:19 <McCool> (also - semantic tagging...)
14:52:27 <plh> dom: comparison with search engines is listed in the AI impact report
14:52:43 <plh> ... whether robots.txt should be used to control AI robots
14:52:53 <danbri> at some point Google started using a new string for AI usecases, https://developers.google.com/search/docs/crawling-indexing/overview-google-crawlers ... not sure if that got rolled back
14:53:12 <plh> .... but search engines operated on prefetch content, while AIs are doing live-fetch content
14:53:39 <plh> ... open sources complaining about AI agents fetching documentation, etc.
14:54:02 <ivan> q?
14:54:08 <plh> ... the more we can anchor ourselves in problems we already know, the better
14:54:15 <ivan> ack RobSmith
14:54:15 <Zakim> RobSmith, you wanted to mention parallels with search engines
14:54:31 <plh> RobSmith: correct about pfretech/live, avoiding DOS attacks
14:54:45 <plh> dom: it's distributed DOS :(
14:54:57 <plh> dom: in conclusion
14:55:04 <danbri> see https://developers.google.com/search/docs/crawling-indexing/google-common-crawlers "Google-Extended is a standalone product token that web publishers can use to manage whether their sites help improve Gemini Apps and Vertex AI generative APIs, including future generations of models that power those products. Grounding with Google Search on
14:55:04 <danbri> Vertex AI does not use web pages for grounding that have disallowed Google-Extended. Google-Extended does not impact a site's inclusion or ranking in Google Search."
14:55:04 <danbri> This is for sites to use robots.txt to exclude crawlers that feed certain AI apps without dropping out of general search (afaik; I am no longer at GOogle and do not speak for them!)
14:55:07 <plh> ... upcoming AC meeting has a session on AI
14:55:31 <plh> ... will look into danbri suggestion on building a broader view from all of our groups to AI impact
14:55:45 <plh> ... not ready to commit to a workshop or a cg, but get in touch with me
14:55:54 <plh> ... very interested in contacts
14:56:08 <plh> ... dom@w3.org
14:56:19 <plh> ... anything else?
14:56:20 <anssik> q+
14:56:54 <plh> anssi: should we aim to create a web platform where you can't distinguish between humans and AIs?
14:57:00 <hadleybeeman> Can we stop it happening?
14:57:19 <plh> dom: also, should we be able to detect it?
14:57:32 <plh> ... very interesting way of framing the debate
14:57:52 <plh> danbri: get and post are very different
14:58:05 <plh> ... more interesting in dinstinguishing the post than the get
14:58:15 <plh> dom: feel free to get in touch with me
14:58:42 <plh> [end of discussion]
14:58:47 <ivan> ivan has left #ai-agents
14:59:01 <dom> RRSAgent, draft minutes
14:59:02 <RRSAgent> I have made the request to generate https://www.w3.org/2025/03/26-ai-agents-minutes.html dom
14:59:46 <dom> i|-> https://www.w3.org/2025/Talks/dhm-ai-agents/ Slides|Slideset: https://www.w3.org/2025/Talks/dhm-ai-agents/
14:59:48 <dom> RRSAgent, draft minutes
14:59:50 <RRSAgent> I have made the request to generate https://www.w3.org/2025/03/26-ai-agents-minutes.html dom
15:01:36 <denis> denis has left #ai-agents
15:02:43 <hyojin> hyojin has left #ai-agents
15:21:15 <arichiv> arichiv has joined #ai-agents
22:58:01 <tidoust> RRSAgent, bye
22:58:01 <RRSAgent> I see no action items