14:52:27 <RRSAgent> RRSAgent has joined #generative-ai
14:52:31 <RRSAgent> logging to https://www.w3.org/2024/09/25-generative-ai-irc
14:52:31 <tpac-breakout-bot> RRSAgent, do not leave
14:52:32 <tpac-breakout-bot> RRSAgent, make logs public
14:52:33 <tpac-breakout-bot> Meeting: Web and Generative AI risks
14:52:33 <tpac-breakout-bot> Chair: Qing An
14:52:33 <tpac-breakout-bot> Agenda: https://github.com/w3c/tpac2024-breakouts/issues/36
14:52:33 <Zakim> Zakim has joined #generative-ai
14:52:34 <tpac-breakout-bot> Zakim, clear agenda
14:52:34 <Zakim> agenda cleared
14:52:34 <tpac-breakout-bot> Zakim, agenda+ Pick a scribe
14:52:36 <Zakim> agendum 1 added
14:52:37 <tpac-breakout-bot> Zakim, agenda+ Reminders: code of conduct, health policies, recorded session policy
14:52:37 <Zakim> agendum 2 added
14:52:37 <tpac-breakout-bot> Zakim, agenda+ Goal of this session
14:52:37 <Zakim> agendum 3 added
14:52:37 <tpac-breakout-bot> Zakim, agenda+ Discussion
14:52:37 <Zakim> agendum 4 added
14:52:37 <tpac-breakout-bot> Zakim, agenda+ Next steps / where discussion continues
14:52:39 <Zakim> agendum 5 added
14:52:39 <tpac-breakout-bot> tpac-breakout-bot has left #generative-ai
15:13:55 <tantek> tantek has joined #generative-ai
15:31:10 <x> x has joined #generative-ai
15:32:29 <Lei_Zhao> Lei_Zhao has joined #generative-ai
15:32:45 <edward> edward has joined #generative-ai
15:34:00 <angel> angel has joined #generative-ai
15:34:40 <angel> zakim, this is Generative AI Breakout session
15:34:40 <Zakim> got it, angel
15:34:46 <chrisp> chrisp has joined #generative-ai
15:34:52 <angel> scribe: angel
15:34:58 <angel> chair: Qing An
15:35:21 <angel> QingAn: now we can start the meeting
15:35:25 <riju> riju has joined #generative-ai
15:35:31 <angel> ...let start with an short intro
15:35:54 <lambda> lambda has joined #generative-ai
15:36:01 <kavramesh> kavramesh has joined #generative-ai
15:36:05 <CharlesL> CharlesL has joined #generative-ai
15:36:25 <angel> ...This session will introduce the new features and new risks brought by generative AI compared with traditional AI. And the session will have an open discussion about the risks related to Web and W3C can do on Web standards to address the risks.
15:36:33 <RobKochman> RobKochman has joined #generative-ai
15:36:38 <angel> ...this session is under W3C Code of conduct
15:36:45 <angel> ...and it will not be recorded
15:36:50 <ivan> ivan has joined #generative-ai
15:36:52 <Yves> Yves has joined #generative-ai
15:36:53 <angel> ...now I will start the intro slides
15:37:37 <angel> ...Generative Artificial Intelligence (AI) is happening everywhere
15:37:46 <angel> ...people is using it in their daily life
15:38:22 <angel> ...many orgs are evaluating risks the Generative Artificial Intelligence (AI)  might bring to people's life and other aspects
15:38:40 <angel> ... Web is an important venue to deliver the AI servive to people
15:38:50 <angel> s/servive /service
15:39:27 <angel> ...first I will go with the features of Generative Artificial Intelligence then the risks it might bring, and discuss with you about possible solutions
15:39:32 <RobKochman> RobKochman has joined #generative-ai
15:40:16 <Igarashi> Igarashi has joined #generative-ai
15:40:17 <angel> ...we have been using the Conventional AI for many years
15:40:26 <Igarashi> present+
15:40:42 <angel> ...there are many mature solutions to handle the risks brought by Conventional AI
15:40:43 <ivan> present+
15:41:09 <tomayac> tomayac has joined #generative-ai
15:41:24 <angel> ...list 5 risks brought by Conventional AI, such as Classification task: to predict discrete labels, Regression task: to predict continuous values, etc
15:41:39 <angel> Qing AN's slides -> https://github.com/user-attachments/files/17092045/Web.and.Generative.AI.risks-Breakout.pdf
15:42:04 <angel> Qing AN: in trandisional AI, one task with one risk to handle
15:42:35 <angel> ...but .Generative Artificial Intelligence has different goal, which is to generate new data instances that are
15:42:35 <angel> statistically consistent with the training data
15:43:20 <angel> ...another difference is the output:  output of Generative Artificial Intelligenc is to get new and complete data
15:43:20 <angel> instances, such as an image or a piece of tex
15:43:52 <angel> ...New features of Generative AI include new contents generated
15:44:19 <kenji_baheux> kenji_baheux has joined #generative-ai
15:44:50 <angel> ...2nd feature is the long context windows and self-attention mechanism, which enables different attention
15:45:07 <angel> ...3rd feature is the low production barrier
15:45:27 <angel> ...Contents are easily generated via natural language conversation.
15:45:39 <angel> ...4th feature is Highly convincing generated contents
15:45:59 <angel> ... and the 5th feature is greater randomness of generated contents
15:46:12 <RK> RK has joined #generative-ai
15:46:23 <angel> ...based on these new features, there are new kinds of risks
15:46:44 <angel> ...risk #1: Eased access to knowledge
15:46:55 <angel> ... Easy access to knowledge might make it easier for malicious users to cause harms to society without specialized training (e.g., CBRN knowledge, malware)
15:47:12 <angel> ...Risk #2: Generalization of input understanding and output generation
15:47:37 <angel> ... Generative AI’s strong generalization ability might result in hallucination, which will cause further harms if users believe the generated false content due to the highly convincing output
15:47:51 <angel> ...Risk #3: Use of generated contents
15:48:06 <angel> ... The generated contents, when contain faults or misalign with regulations and ethics, might mislead the downstream applications make incorrect decisions or even harmful actions.
15:48:18 <angel> ...the outcome can be serious is the application is about heath care
15:48:36 <angel> ...Risk #4: Influence of generated contents
15:48:47 <angel> ... Generative AI might generate unethical contents that pose harms to individuals and society
15:48:58 <angel> ...on one hand, Generative AI might generate disrespectful, dangerous contents that promote self-harm or illegal activities
15:49:16 <angel> ...on the other hand, Also, it might generate biased output, which can cause unfair distribution of benefits from using the generative AI (e.g., the quality of generated contents may be better for some groups but worse for others, due to different input format or language)
15:49:51 <angel> ...some research shows if the content is training on English content only, it might have certain bias and create unfair results
15:50:18 <angel> ... Risk #5: Human reliance on generative AI
15:50:31 <angel> ... Over-reliance on generative AI might cause humans to be manipulated, especially when humans have no detailed knowledge of how generative AI works
15:51:03 <angel> ... Risk #6: Privacy
15:51:14 <angel> ... Due to training data memorization, the generated contents might cause sensitive information leakage
15:51:23 <angel> ...While users can benefit from the customized personal assistant by feeding personal information to generative AI.
15:51:38 <angel> ...risk #7: Copyright and Intellectual Property
15:51:47 <angel> ... Due to training data memorization, the generated contents might cause copyrights infringement.
15:52:00 <xueyuan> xueyuan has joined #generative-ai
15:52:01 <angel> ...risk #8: Continuous improvement based on reinforced learning
15:52:10 <angel> ... Continuous learning based on the user feedback can be leveraged to mislead generative AI behaviors
15:52:20 <angel> ... also, it can enable better alignment with human preference
15:52:45 <angel> ...risk #9: New security attacks, which mainly about Prompt-based attacks expand the attack surface
15:53:01 <angel> ... about what Web can do to address risks
15:53:26 <angel> ...my personal thinking is that, first is to detecting the malware (hacking, malware, and phishing) generated by Generative A
15:53:38 <angel> ...so no harm to users
15:53:59 <angel> ...one scenario is that user interacting with Generative AI application (typically chat) in Web Browser, malicious Generative AI application might generate malware in chat box and induce users to open the malware
15:54:10 <angel> ... How to have Web Browser support detecting the malware generated by Generative AI in chat box?
15:54:19 <angel> ...I believe we can do something in the Web
15:54:38 <angel> ...the 2nd thing we can do is to increase the accountability
15:55:06 <angel> ...by labelling the output from Generative AI, including images, videos, audios, text and code
15:55:20 <angel> ...which can be done by using explicit watermarking
15:55:39 <angel> ... on the Generative AI  output
15:55:46 <xfq> xfq has joined #generative-ai
15:56:07 <angel> ...one thing the Web can do, is to make the Web Browser recognize the watermarking formats
15:56:51 <angel> ...to make it easier for users to find out if the outcome is AI generated
15:57:06 <angel> ... also, to address the Transparency issue
15:57:18 <angel> ... need to expose the info of AI models
15:57:37 <angel> ...there are some regulations about this to require the AI models to expose the model info
15:57:59 <angel> ... to enhance the AI transparency
15:58:10 <angel> ...and Web is a good platform to deliver this kind of info
15:58:34 <angel> ...to make the users and developers to have easier access to info like this
15:58:48 <angel> ...which helps to increase the public trust of Generative AI,
15:59:07 <angel> ... about Privacy, need secure access to privacy info in local generative AI
15:59:32 <angel> ...need a secure area to store the user privacy info
16:00:11 <angel> ...would it be benefitial to develope an API to enable securely access the user privacy info
16:00:17 <angel> ...above are my thoughts
16:00:41 <angel> ...I hope we can have good disucsions about if these are reasonable directions we should be chasing in W3C
16:00:44 <angel> ...comments are welcome
16:00:47 <angel> q?
16:01:17 <Vanitha> Vanitha has joined #generative-ai
16:01:48 <QingAN> QingAN has joined #generative-ai
16:01:50 <Igarashi> q?
16:01:54 <Igarashi> q+
16:01:57 <kenji_baheux> q+
16:02:03 <angel> ack igara
16:02:05 <QingAN> q?
16:02:28 <angel> Igarashi san: one qestion about the standardization of generative AI
16:03:46 <angel> ...do you know any other SDO working on this kind of data model standards?
16:04:10 <angel> another question is about what kind of AI you are thinking about to protect the privacy?
16:04:34 <angel> ...JS API or API between the client and service?
16:05:02 <angel> Qing AN: CSA is working on the this topic
16:05:18 <Vanitha> Vanitha has joined #generative-ai
16:05:29 <angel> ...they have published some reports to define some data suite, which might be called model suite
16:05:38 <jfernandez> jfernandez has joined #generative-ai
16:05:56 <angel> ...they are working on what kind of @@1 to inlcude what kind of data models
16:06:02 <angel> ...but not about the Web
16:06:12 <angel> ...that why I think W3C can work on it
16:06:32 <angel> ...for the 2nd question, I agree it is not quite related to the browser, more on the AI side
16:06:58 <angel> ...I think there is another breakout session on this topic today
16:06:58 <QingAN> q?
16:07:04 <angel> ack Kenji
16:07:27 <angel> Kenji san: from google, if you have looked into C2PA
16:07:43 <kenji_baheux> https://c2pa.org/specifications/specifications/2.0/specs/C2PA_Specification.html
16:07:53 <angel> QingAI: I think it is one commonly used solution to achieve the watermarking
16:08:21 <angel> ...W3C does not need to reinvite the wheels
16:08:40 <vmpstr> vmpstr has joined #generative-ai
16:08:42 <angel> ...if we are using the explicit wather marking, it is not very obvious for the users to see it
16:09:09 <angel> if making the explicit wather marking obvious, it might also affect the UX
16:09:16 <Igarashi> both C2PA approach(digital cert.) and watermaking could be used, but no standard watermaking, as I know.
16:09:38 <angel> ...I think what the web can do is to expose what kind of watermarking it is using
16:09:48 <cwilso> cwilso has joined #generative-ai
16:09:50 <angel> ...make it easier fo the user to get this kind of info
16:09:53 <cwilso> present+
16:09:53 <angel> q?
16:10:01 <QingAN> q?
16:10:14 <ivan> s/water marking/watermarking/g
16:11:04 <angel> QingAN: on the zoom, anyone on the queue?
16:11:07 <ivan> s/wather marking/watermarking/g
16:11:32 <angel> Chris Pryor: @@2
16:11:37 <QingAN> q?
16:12:14 <angel> QingAn: if we want to do the next step? where we should start the work in W3C?
16:12:24 <edward> edward has joined #generative-ai
16:13:49 <angel> ...I think the first step is to to call for participants for this topic
16:14:02 <angel> ... see if we can have some experts with common interests
16:14:18 <angel> ...and discuss if there is any topics worth further efforts from W3C
16:14:31 <angel> ...if there is, maybe we can set up a CG to future incubate this idea
16:15:58 <angel> ...next I will tlak to W3C team and see if we can find any proper way to move forward this topic
16:16:07 <angel> s/tlak/talk
16:16:24 <angel> ...and maybe set up a CG to have futher disusion
16:16:33 <angel> s/disusion/discussion
16:17:12 <ivan> ivan has left #generative-ai
16:17:19 <angel> ...suggested summary: proprose to discuss with W3C team and prepare for settiing up a CG for this topic
16:18:13 <angel> rrsagent, make log
16:18:13 <RRSAgent> I'm logging. I don't understand 'make log', angel.  Try /msg RRSAgent help
16:18:24 <angel> rrsagent, draft minutes
16:18:26 <RRSAgent> I have made the request to generate https://www.w3.org/2024/09/25-generative-ai-minutes.html angel
16:18:33 <angel> rrsagent, make log public
16:18:42 <QingAN> This is another related session I mentioned: https://github.com/w3c/tpac2024-breakouts/issues/44
16:18:58 <angel> rrsagent, draft minutes
16:18:59 <RRSAgent> I have made the request to generate https://www.w3.org/2024/09/25-generative-ai-minutes.html angel
16:29:40 <xfq> xfq has left #generative-ai
17:04:19 <Yves> Yves has left #generative-ai
17:05:54 <RobKochman> RobKochman has joined #generative-ai
17:24:25 <jfernandez> jfernandez has left #generative-ai
17:28:30 <RobKochman> RobKochman has left #generative-ai