(ℹ) Join us now at the IRC channel | ䷉ Find the plain text version at this address (HTTP) or in Gemini (how to use Gemini) with a full GemText version.
schestowitz[TR3] | https://lists.gnu.org/archive/html/libreplanet-discuss/2025-07/msg00028.html | Jul 23 02:54 |
---|---|---|
-TechBytesBot/#techbytes-lists.gnu.org | Re: Is AI-generated code changing free software? | Jul 23 02:54 | |
schestowitz[TR3] | "Hi Jean Louis, et al, | Jul 23 02:54 |
schestowitz[TR3] | I guess I didn't make myself clear. Configuration/Setup is not the issue. The Numen project (in combination with 'dotool'/'xdotool' and 'keynav') provides a complete voice computer control system. Configuration is via a set of text files and is surprisingly flexible. I now have custom setup which, in principle, allows me to easily do all the tasks that you mention in your reply and many more. Basically, I should now be able to | Jul 23 02:54 |
schestowitz[TR3] | do with my voice anything that I could do with the mouse and keyboard. If the underlying AI/LLM speech recognition system was up to the task this setup would be awesome! Unfortunately, it's not. | Jul 23 02:54 |
schestowitz[TR3] | Numen uses Vosk (https://alphacephei.com/vosk/ ) as it's speech recognition engine. (I believe Vosk is in turn based on the Kaldi speech recognition toolkit: https://kaldi-asr.org/doc/about.html .) According to the documentation for Numen and Vosk, the default model used by Numen is 'vosk-model-small-en-us-0.15' which has a word error rate (WER) of around 10 (see https://alphacephei.com/vosk/models ). This means that, on average | Jul 23 02:54 |
schestowitz[TR3] | , around one word ten will be wrong. | Jul 23 02:54 |
schestowitz[TR3] | At first glance, a WER of 10 doesn't sound too bad for dictation ... but consider: Imagine that you've just dictated an e-mail with 200 words in it. About 20 of those will be wrong. That means not only that one has to go back and proof read the e-mail to look for the errors, but, presumably, voice control will _also_ be used to correct the errors also. But some of those commands will be misunderstood creating yet more problems. | Jul 23 02:54 |
schestowitz[TR3] | Which brings me to ... | Jul 23 02:54 |
schestowitz[TR3] | For computer control functions (editing text, selecting between application windows, selecting menu items, clicking on buttons, changing work spaces, etc., etc.) is sometimes a complete mess. In this case the AI/LLM is frequently being used to enter various control sequences. When _this_ goes wrong it can be a complete disaster! It can (unintentionally) delete big chunks of text, delete large numbers of emails, close application | Jul 23 02:54 |
schestowitz[TR3] | windows, put the keyboard and/or mouse in an unresponsive state, and so on. Recovering from such errors has sometimes taken me an hour or more. That certainly doesn't do much for my productivity. :^/ | Jul 23 02:54 |
schestowitz[TR3] | Clearly there are cases where even this level of functionality/reliability would be a huge win. Somebody who is unable to use a keyboard and mouse at all for example. For me, as I've said, it's basically a draw. For somebody with no problem using a mouse and keyboard it would just be a huge PITA. | Jul 23 02:54 |
schestowitz[TR3] | As always though, this is just my $0.02." | Jul 23 02:54 |
-TechBytesBot/#techbytes-alphacephei.com | VOSK Offline Speech Recognition API | Jul 23 02:54 | |
-TechBytesBot/#techbytes-kaldi-asr.org | Kaldi: About the Kaldi project | Jul 23 02:54 | |
-TechBytesBot/#techbytes-alphacephei.com | VOSK Models | Jul 23 02:54 | |
schestowitz[TR3] | <li> | Jul 23 07:50 |
schestowitz[TR3] | <h5><a href="https://linuxiac.com/google-debuts-oss-rebuild-project/">Google Debuts OSS Rebuild Project</a></h5> | Jul 23 07:50 |
schestowitz[TR3] | <blockquote> | Jul 23 07:50 |
schestowitz[TR3] | <p>In simple terms, it attempts to rebuild what developers download, verify that the binaries originated from the public source tree, and raise an alarm if anything appears suspicious. Here’s how the whole thing works. </p> | Jul 23 07:50 |
schestowitz[TR3] | </blockquote> | Jul 23 07:50 |
schestowitz[TR3] | </li> | Jul 23 07:50 |
-TechBytesBot/#techbytes-linuxiac.com | Google Debuts OSS Rebuild Project | Jul 23 07:50 | |
*psydruid (~psydruid@jevhxkzmtrbww.irc) has left #techbytes | Jul 23 07:53 | |
schestowitz[TR3] | <li> | Jul 23 07:55 |
schestowitz[TR3] | <h5><a href="https://linuxiac.com/fwupd-2-0-13-released-with-faster-startup-and-lower-memory-use/">Fwupd 2.0.13 Released with Faster Startup and Lower Memory Use</a></h5> | Jul 23 07:55 |
schestowitz[TR3] | <blockquote> | Jul 23 07:55 |
schestowitz[TR3] | <p>Over a month after its previous 2.0.12 release, Fwupd, an open-source utility designed to simplify firmware updates on Linux-based systems, has rolled out its new 2.0.13 version. </p> | Jul 23 07:55 |
schestowitz[TR3] | </blockquote> | Jul 23 07:55 |
schestowitz[TR3] | </li> | Jul 23 07:55 |
-TechBytesBot/#techbytes-linuxiac.com | Fwupd 2.0.13 Released with Faster Startup and Lower Memory Use | Jul 23 07:55 | |
*psydruid (~psydruid@jevhxkzmtrbww.irc) has joined #techbytes | Jul 23 08:08 | |
schestowitz[TR3] | He workeHe whttp://ipkitten.blogspot.com/2025/07/peas-and-peculiarities-of-product-by.html?showComment=1753091007461#c7573737861505723230 | Jul 23 12:57 |
schestowitz[TR3] | "I've found with antibody cases that the burden of proof stays with the patentee in opposition for showing novelty/inventive step for complex definitions of binding properties, though the Guidelines does not explicitly state it should. So with product by processes I believe the burden should stay with patentees in opposition, especially where the process imparts complex structural" | Jul 23 12:57 |
*psydruid (~psydruid@jevhxkzmtrbww.irc) has left #techbytes | Jul 23 15:35 | |
*psydruid (~psydruid@jevhxkzmtrbww.irc) has joined #techbytes | Jul 23 16:32 | |
*psydroid3 (~psydroid@yu29f4abyrsnc.irc) has joined #techbytes | Jul 23 18:08 | |
*psydroid3 has quit (Quit: KVIrc 5.2.6 Quasar http://www.kvirc.net/) | Jul 23 22:36 |
Generated by irclog2html.py
2.6 | ䷉ find the plain text version at this address (HTTP) or in Gemini (how to use Gemini) with a full GemText version.