Ivan Herman

Retired; emeritus W3C team member, still active in digital publishing, semantic web, data, verifiable credentials; also interested in history, arts (painting, sculpture, architecture), photography, international politics, literature, music (mostly classical).

Public Key

npub13875z6etm6q5f2e4wwvf2dzjv2kfn6whn9vxvman9tqg394lx2nsf24xgx

NIP-05 Address

ivan_herman_at_w3c.social@mostr.pub

Profile Code

nprofile1qqsgnl2pdv4aaq2y4v6h8xy4x3fx9tyea8tejkrxd7ej4sygj6ln9fcpz4mhxue69uhhyetvv9ujumt0wd68ytnsw43qytqqvx

Publishing to

relay.mostr.pub

Author Public Key

npub13875z6etm6q5f2e4wwvf2dzjv2kfn6whn9vxvman9tqg394lx2nsf24xgx

Show more details

Last Notes

2024-09-04 09:36:26

Have not tried it yet, but good to know it exists:
"ElevenLabs' text-to-speech app Reader is now available globally"
https://techcrunch.com/2024/08/19/elevenlabs-reader-app-is-now-available-globally/

2024-08-17 13:19:03

[2408.07892] Personhood credentials: Artificial intelligence and the value of privacy-preserving tools to distinguish who is real online
https://arxiv.org/abs/2408.07892

2024-08-11 08:57:35

Thanks to @npub1ftd…4cve for pointing at this in an article he co-authored.
https://www.promarket.org/2024/08/08/chrome-is-the-forgotten-fulcrum-of-googles-dominance/

2024-08-01 01:59:48

https://blog.1password.com/ai-browser-extension-nightmare/

2024-07-26 04:30:44

https://www.w3.org/2001/tag/doc/web-without-3p-cookies/

2024-07-24 13:30:35

"Would you believe that the entire ebook marketplace – including Kindle, iBooks, and thousands of ebook stores – depends on the volunteer work of about a dozen people?"
(interview with Dave Cramer)
https://howbooksaremade.com/episodes/how-do-ebooks-work/
#ebooks #epub

2024-07-11 13:17:36

US officials announce the takedown of an AI-powered Russian bot farm
https://www.engadget.com/us-officials-announce-the-takedown-of-an-ai-powered-russian-bot-farm-054034912.html
#ai #politics #russia

2024-07-11 13:08:03

IAB Workshop on AI-CONTROL (aicontrolws)
https://datatracker.ietf.org/group/aicontrolws/about/
#ai

2024-06-28 07:31:11

https://www.engadget.com/please-dont-get-your-news-from-ai-chatbots-000027227.html #ai

2024-06-24 15:16:27

How to Fix “AI’s Original Sin” – O’Reilly
https://www.oreilly.com/radar/how-to-fix-ais-original-sin/
#ai #copyright

2024-05-28 10:50:39

This may become an important approach in practice: identifiers that are resistant to minor changes in the media, ie, closer to identify the "abstract" content itself.
https://iscc.io/iso-publishes-new-standard-enabling-content-transparency-iso-24138/

2024-05-21 09:53:34

A long overdue standard has now been published at W3C: "RDF Dataset Canonicalization". Essential for security related operations with RDF Datasets...
Note that the standard is published alongside 8 full implementations, which is great.
https://www.w3.org/news/2024/rdf-dataset-canonicalization-is-a-w3c-recommendation/
#rdf

2024-05-08 04:54:38

- reply

@npub184l…3acz It is the same attitude of the ultra right from Hungary to France, from Italy to the Netherlands...

2024-04-25 04:39:57

- reply

@npub1y4n…25mf It is obviously, not perfect, but I didn't hit any glaring issues so far. I didn't do any systematic tests though.

2024-04-23 09:25:29

Very much worth reading.
"The internet has become an extractive and fragile monoculture. But we can revitalize it using lessons learned by ecologists"
https://www.noemamag.com/we-need-to-rewild-the-internet/

2024-04-22 14:23:53

I enjoyed reading this book; also learned a lot.
https://www.penguinrandomhouse.com/books/690603/tracers-in-the-dark-by-andy-greenberg/

2024-04-19 16:40:36

- reply

@npub1tlh…8myw I did some further comparisons and I found that perplexity seems to be better when using French. Would be interesting in general to compare these on language axes...

2024-04-19 06:28:24

Open Source Security (OpenSSF) and OpenJS Foundations Issue Alert for Social Engineering Takeovers of Open Source Projects
https://openjsf.org/blog/openssf-openjs-alert-social-engineering-takeovers
#security

2024-04-17 09:42:02

I have been using the perplexity.ai service for a while now, and I really like it. It is an LLM based conversational interface to Web search, which has the added quality of putting exact references into its response, so you can go back to the “real” sources. This “minor” touch makes it more useful than the other LLM based services in my view…
Of course, all answers are to be taken with a pinch of salt, but that is true for any web search after all...
https://www.perplexity.ai
#ai

2024-04-03 06:50:23

An unending array of jailbreaking attacks could be the death of LLMs. The conclusion:
"Nobody knows exactly how LLMs work, and that means that nobody can ever issue strong guarantees around them. That would be fine if LLMs were kept in the lab, where they arguably still belong, but with hundreds of millions of people using them daily […] the lack of any kind of guarantee gets more worrisome by the day."
https://garymarcus.substack.com/p/an-unending-array-of-jailbreaking

2024-03-31 16:05:26

A Terse Profile for JSON-LD. May be an interesting development; a simplified view of JSON-LD.
https://zenomt.github.io/jsonld-terse/
#rdf #jsonld

2024-03-31 08:15:49

(AI) Models All The Way Down
Nice overview on the image training sets, how the training sets actually depend on other AI models and not (necessarily) by humans. Meaning that bias in one model infects other models, and then again...
(To view the text, you have to scroll down in the browser.)
https://knowingmachines.org/models-all-the-way
#ai

2024-03-30 11:45:39

- reply

@npub1ffq…4kpt The homebrew blog page only refers to a version of Dec '23 as part of the announcement. Isn't that the version that introduced the bug?

2024-03-30 11:24:36

- reply

Actually... the article refers to a usage pattern for "those suffering from sudden or degenerative speech conditions." Having gone through a similar ordeal lately, I am not sure how such an application would have helped me. It sounds like a technology that is not necessarily need, done "just because we can".
(Of course, there are plenty of dangerous applications out there: we will get totally realistic spams or fake news...)

2024-03-21 10:44:20

"Counterfeit Books Are a Growing Problem, but a Solution May Be Near"
https://www.publishersweekly.com/pw/by-topic/industry-news/publisher-news/article/94446-fake-books-are-a-real-problem-but-a-solution-may-be-near.html
#publishing #epub

2024-03-16 04:56:33

The exponential enshittification of science
"…there is no way reviewers and journals are going to be able to keep up. Reviewers are typically unpaid academics who are already stretched to their limits; tripling their workload would not be feasible. […] the total number of articles may radically spike, many of them dubious and a waste of reviewers’ time. Lots of bad stuff is going to sneak in."
https://garymarcus.substack.com/p/the-exponential-enshittification
#ai #science #publishing

2024-03-04 15:35:25

- reply

@npub1dzv…mt3x It looks perfectly fine to me.
I run the latest version of Arc (v1.32.0).

2024-02-25 11:23:47

"Google has apologized (or come very close to apologizing) for another embarrassing AI blunder this week, an image-generating model that injected diversity into pictures with a farcical disregard for historical context. While the underlying issue is perfectly understandable, Google blames the model for “becoming” oversensitive. But the model didn’t make itself, guys."
https://techcrunch.com/2024/02/23/embarrassing-and-wrong-google-admits-it-lost-control-of-image-generating-ai/

2024-02-06 15:57:51

https://techcrunch.com/2024/02/05/yandex-to-sell-its-remaining-russian-businesses-for-5-2b-half-its-market-value/

2024-02-04 11:56:35

A new direction for Web search (you use it via a web browser, or a web app installed locally...).
https://blog.perplexity.ai/blog/perplexity-raises-series-b-funding-round
#ai

2024-02-03 16:11:37

The (new) "browse for me" feature of the arc browser is neat. Type in a search term, the browser goes through the search, takes the top hits, looks at their content, and synthesizes the results into a readable answer (I presume some AI tool is behind this). The sources are clearly mentioned, and the good-old search is still possible; this is just a new feature. So far it has proven to be pretty helpful. And a good idea, for sure.
https://arc.net/
#ArcBrowser

2024-01-30 09:55:17

https://www.theverge.com/2024/1/24/24049047/data-center-ai-crypto-bitcoin-mining-electricity-report-iea

2024-01-30 07:44:58

"[…] the evidence to date suggests that there is real potential for generative AI to improve the accuracy of medical diagnoses[…]. Well before the adjunctive use of AI was considered, there was ample evidence that physician biases contributed to medical diagnostic errors, such as. […] It is easy to forget that no physician can possibly keep up with all the medical literature on the approximately 10,000 types of human disease."
https://erictopol.substack.com/p/toward-the-eradication-of-medical
#ai

2024-01-28 09:25:20

- reply

@npub1fp6…wy2j
A totally different comment (not suprising coming from me I guess): you should try to join the W3C Publishing CG (https://www.w3.org/community/publishingcg/). I think this could lead to an interesting discussion with other EPUB people (at least I hope), and give more visibility to your ideas.

2024-01-27 08:25:20

Huge changes for us, EU citizens: alternative app stores, non-webkit browsers, etc. https://www.macrumors.com/2024/01/25/ios-17-4-beta-1-changes/

2024-01-24 15:16:16

Everything you always wanted to know about RDF literals (and you did not dare to ask😀)
https://www.youtube.com/watch?v=FTeZUDsKxqA
#rdf #semanticweb

2024-01-10 07:13:02

On C2PA and photography from Tim Bray. Good to read...
https://www.tbray.org/ongoing/When/202x/2023/10/28/C2PA-Workflows

2024-01-06 17:56:01

I had some fun with this "Obsidian" tool/toy. Standard markdown based note-taking application, can share notes with the apps running on my iPad and iPhone by simply storing them on iCloud, can dump one, or possibly all, notes in HTML. It has a plugin architecture with a bunch of community plugins (although I found a bug in one of them...). It is fully functional for private use free of charge. Good stuff.
https://obsidian.md/

2023-12-17 10:35:47

With elections taking place in 2024 in the US, in Europe, and elsewhere…
“AI chatbot got election info wrong 30 percent of time.”
Some other quotes:
“[…]AI chatbots could contribute to confusion and misinformation around future elections […] giving the wrong date for elections, reporting outdated or mistaken polling numbers, listing candidates who had withdrawn from the race as leading contenders, and inventing controversies about candidates in a few cases."
https://www.washingtonpost.com/technology/2023/12/15/microsoft-copilot-bing-ai-hallucinations-elections/
#ai

2023-12-08 19:53:27

- reply

@npub14ux…dwn4 @npub1w7a…tpn3 I don’t know the exact origin either. Another story I did hear is to choose a name for a specification that is all about logic something that is illogical… Note that these two stories don’t exclude one another.

2023-12-07 10:53:12

Pour les francophones entre nous…
Qu'on soit croyant ou non, un épisode très intéressant du podcast "Le code a Changé" : "Un catholique face à IA".
https://www.radiofrance.fr/franceinter/podcasts/le-code-a-change/le-code-a-change-4-5957439

2023-11-29 09:00:17

Have OpenAI made an AGI breakthrough? https://webscience.org/have-openai-made-an-agi-breakthrough/

2023-11-19 14:52:31

““Godfather of AI” Geoff Hinton, in recent public talks, explains that one of the greatest risks is not that chatbots will become super-intelligent, but that they will generate text that is super-persuasive without being intelligent, in the manner of Donald Trump or Boris Johnson. “
From “Oops! We automated Bullshit.” from Alan Blackwell
https://www.cst.cam.ac.uk/blog/afb21/oops-we-automated-bullshit

2023-11-08 07:32:58

“Some gen AI vendors say they’ll defend customers from IP lawsuits. Others, not so much.”
https://techcrunch.com/2023/10/06/some-gen-ai-vendors-say-theyll-defend-customers-from-ip-lawsuits-others-not-so-much/?guccounter=1
#ai #copyright

2023-11-07 05:07:19

Sigh. But not surprising. It was bound to happen... https://www.cnn.com/2023/11/02/tech/microsoft-ai-news/index.html

2023-11-06 16:40:24

“Lawrence Lessig on why AI and social media are causing a free speech crisis for the internet”
https://www.theverge.com/23929233/lawrence-lessig-free-speech-first-amendment-ai-content-moderation-decoder-interview

2023-11-02 05:28:04

Summit on AI at one of the birthplaces of computing...
https://www.theguardian.com/technology/2023/nov/01/uk-us-eu-and-china-sign-declaration-of-ais-catastrophic-danger

2023-10-26 10:25:18

“The purpose of the Open UI, […], is to allow web developers to style and extend built-in web UI components and controls, such as <select> dropdowns, checkboxes, radio buttonss […].
To do that, we’ll need to fully specify the component parts, states, and behaviors of the built-in controls, as well as necessary accessibility requirements, and provide test suites to ensure compatibility. We’ll also implement polyfills for our extensible web UI controls.”
https://open-ui.org
#web

2023-10-21 07:39:57

Connie Chan’s Internet Archive Hero Award 2023 speech (extracts), on the value of free library and information archive…
https://archive.org/details/connie-chan-hero-award

2023-10-20 07:42:08

https://www.engadget.com/the-eu-is-probing-meta-and-tiktoks-israel-hamas-response-170057848.html?src=rss

2023-10-15 14:01:27

- reply

@npub14ux…dwn4 @npub19mz…d86c At least for me, I authenticate with my fingerprint on my mac or iPad, and face ID for iPhone when I use 1Password for passkey. Of course it is true that, once in a while, 1P asks for a password (but that isn't bound to passkey usage).

2023-10-06 10:03:57

Anybody using the “keyword search” safari extension on safari 17 on sonoma? It does not seem to work for me since I upgraded, nor does a newly installed similar extension work. Has safari removed an event these extensions rely on?

2023-10-04 12:51:59

- reply

@npub10hr…njzx Chrome/Chromium also had tab groups and, at least Brave, profiles. But there are some major deficiencies: e.g., it is not clear how (if at all) I can ‘save’ tab groups; they are lost if you quit the browser. As for profiles: it does not seem to work at all, at least not on the latest MacOS…

2023-09-29 12:43:54

Interesting experiment on the same text being read by the author, and some voices generated by AI. See some more explanation in the caption of the video (well, essentially a series of audio clips).
https://www.youtube.com/watch?v=910I5aBp8q8
#ai
#tts

2023-09-22 10:04:42

“La Chine sous l’emprise de l’intelligence artificielle” - (in French) from France24. The real meat is a video roughly between 1:57 and 13:50.
https://www.youtube.com/watch?v=JabCQ6Y_M5s
#AI

2023-09-21 08:40:27

This may have serious consequences (and it is not the only one of the sort):
“The Authors Guild and 17 authors filed a class-action suit against OpenAI in the Southern District of New York for copyright infringement of their works of fiction on behalf of a class of fiction writers whose works have been used to train GPT. ”
#generativeAI #chatgpt
See details in https://authorsguild.org/news/ag-and-authors-file-class-action-suit-against-openai/

2023-09-05 04:38:38

- reply

@npub1698…hmgt @npub1xq6…lysw Never looked at Omnigraffle, to be honest. I am not against paying software in principle, but for most of the features there are perfectly fine free alternatives…

2023-09-04 15:06:09

- reply

@npub1xq6…lysw I use draw.io which works pretty well and exports very decent SVG.

2023-09-04 14:36:16

- reply

@npub1xq6…lysw
3/ But my biggest problem is unpredictability. Dark mode is either a choice of the OS or the browser that enforces dark mode on Web content. I presume there is an automatic transformation of the colour space in some ways, and the tools do it differently (no standard…).
I would love to see an extension of the colour contrast checking tools that would also show the effect of the same colour combination after imposing dark mode. That would certainly help…

2023-09-04 14:31:44

- reply

@npub1xq6…lysw
2/ The reliance of CSS is only possible with tools that can really use CSS elements. I believe that is the case with Inkscape; but I personally had major issues with Inkscape (it has a very buggy implementation of connecting lines).

2023-09-04 14:31:11

- reply

@npub1xq6…lysw
1/ By “removing fonts” you mean remove text and turn the characters into path-s? That certainly helps, but it is harmful for accessibility, which is one of SVG's strength. B.tw., some drawing tools do that automatically (I believe that is the case with Google's drawing tool).

2023-08-26 13:06:39

Designing for dark mode === dark art 😞
I know there is a way with CSS. But if one generates SVG from a tool (google draw, draw.io, whatever) and wants the result to be used in its original form (e.g., light mode) as well as when the browser forces dark mode: it becomes a royal pain. The result depends on browser, plugin, OS...
Are there "standards" on how browsers should do that? Should there be? How can one prepare for this?