In an absolutely shocking development, the Google Gemini "strategic insights" were dreadful, uninteresting and wrong.
And not just because of unexpected incidents late on. It was just generally awful.
In an absolutely shocking development, the Google Gemini "strategic insights" were dreadful, uninteresting and wrong.
And not just because of unexpected incidents late on. It was just generally awful.
iNaturalist published a couple surveys about their social media presence (Google Forms, obviously). I used them to tell them again what I think about the way they handled communication about their genAI demo. (I also forgot to mention their absence from the Fediverse, but the Google grant is a more pressing issue.) Ok, it's about time I gather the courage to write in the forum, or send a mail.
[Why is everyone and their pets using iNat for their projects (we're in Europe), and why do Observation.org apps need Google position tracking to work?! ๐ซค]
#iNaturalist #genAI #google #gemini #geminiAI #llm #CitizenScience #deGoogle #BigTech
๐ง #DeepMind ha presentato la #FACTS Benchmark Suite, un nuovo strumento per valutare lโaccuratezza delle risposte dei #LLM rispetto ai fatti.
๐ I dettagli: https://www.linkedin.com/posts/alessiopomaro_deepmind-facts-llm-activity-7405573021559775235-tTyD
___
โ๏ธ ๐ฆ๐ฒ ๐๐๐ผ๐ถ ๐ฟ๐ถ๐บ๐ฎ๐ป๐ฒ๐ฟ๐ฒ ๐ฎ๐ด๐ด๐ถ๐ผ๐ฟ๐ป๐ฎ๐๐ผ/๐ฎ ๐๐ ๐พ๐๐ฒ๐๐๐ฒ ๐๐ฒ๐บ๐ฎ๐๐ถ๐ฐ๐ต๐ฒ, ๐ถ๐๐ฐ๐ฟ๐ถ๐๐ถ๐๐ถ ๐ฎ๐น๐น๐ฎ ๐บ๐ถ๐ฎ ๐ป๐ฒ๐๐๐น๐ฒ๐๐๐ฒ๐ฟ: https://bit.ly/newsletter-alessiopomaro
To my mind, LLMs/genAI is simply irredeemably ugly. It is pure brute force. There is no theory or formal framework behind it: throw enough training data at a model and you can force it to learn anything, in as far as learning means forcing the direction of guessing.
The proponents will point at all the cleverness that has gone in to the model architecture and statistical techniques, but then that goes for a nuclear weapon as well. It's still ugly.
Just yesterday I finished @alex and @emilymbender magnificent book, The #AI Con.
I love the last chapter and read it carefully. It will be useful to me personally at work and with my family.
There are two recent examples I would like to share, and today I would react differently.
1. My parents gave my daughters a children's book about inclusivity in the workplace. However, all the illustrations reeked of #GenAI.
I didn't say anything to them at the time, but I've...
Of course, I prefer my own take on the topic discussed in the Depeche Mode song... โ๐
https://mastodon.gamedev.place/@jay/115463770714419247
Coincidently, I generated my song "Essence" shortly before the "Plur1bus" streaming TV show was released, and that show covers a similar topic about individuality / essence.
#DepecheMode #DM #InTheEnd #Plur1bus #Pluribus #Jaycosm #Essence #Music #GenAI #Suno #GoogleFlow #Kdenlive #Intel #OpenVINO #AI
Apparently Amazon puts so little value on its Prime Original series that they can't even be bothered to write summaries about the episodes. Instead, they were getting GenAI to do it.
And it got stuff wrong. Obviously.
I feel sorry for the production crew. Imagine putting in all that work and creativity only to be told that people watching it will be basing their expectations and understanding on a soulless, generic, averaged description of your story line ๐ฌ
Struggling to make #GenAI useful with your private data? Lokesh Chenta shows how RAG โ powered by tools like @LangChainAI & @llama_index โ cuts hallucinations and injects precision. Planning Enterprise Q&A or search?
Start here: https://javapro.io/2025/12/10/solutions-to-achieve-desired-results-from-generative-ai-models/
๐ง #Google ha introdotto lโInteractions API, una nuova interfaccia progettata per abilitare applicazioni #AI avanzate basate su agenti.
๐ Dettagli e Colab per provarla: https://www.linkedin.com/posts/alessiopomaro_google-ai-ai-activity-7405502810177400832-VFa-
___
โ๏ธ ๐ฆ๐ฒ ๐๐๐ผ๐ถ ๐ฟ๐ถ๐บ๐ฎ๐ป๐ฒ๐ฟ๐ฒ ๐ฎ๐ด๐ด๐ถ๐ผ๐ฟ๐ป๐ฎ๐๐ผ/๐ฎ ๐๐ ๐พ๐๐ฒ๐๐๐ฒ ๐๐ฒ๐บ๐ฎ๐๐ถ๐ฐ๐ต๐ฒ, ๐ถ๐๐ฐ๐ฟ๐ถ๐๐ถ๐๐ถ ๐ฎ๐น๐น๐ฎ ๐บ๐ถ๐ฎ ๐ป๐ฒ๐๐๐น๐ฒ๐๐๐ฒ๐ฟ: https://bit.ly/newsletter-alessiopomaro
Caught up on Pluribus, which is my favorite show lately. This is the best kind of science fiction: thoughtful, and makes you think. Utopian and dystopian at once. Examining/exploring humanity through jarring cognitive estrangement. Something that keeps occuring to meโฆ #scienceFiction #tv #genAI
The Bet On Juniors Just Got Better
https://tidyfirst.substack.com/p/the-bet-on-juniors-just-got-better
I did a quick A/B test. Using the APIs, I gave a prompt to both #Gemini 2.5 Flash and #gpt4o. While there wasn't an enormous difference between the results, it was enough: Gemini's response was more thoughtful, detailed, and fit-to-purpose.
Gemini is also (literally) an order of magnitude cheaper (for now, at least).
Clear winner: #Google Gemini
#AI #genAI #vibeCoding #API
#agenticAI (stupidest term ever coined)
#coding #programming
Briefing: Information manipulation in the age of generative artificial intelligence | Think Tank | European Parliament #genai
https://www.europarl.europa.eu/thinktank/en/document/EPRS_BRI(2025)779259
Useful patterns for building HTML tools
https://simonwillison.net/2025/Dec/10/html-tools/#atom-everything
Victoria Strauss on Writer Beware, about amazon's latest rights grab: the AskThis Book feature, rolled out without notification to publishers or authors, and likely infringing on copyright.