Masthash

#ml

Joseph
7 minutes ago

As a CTO, I'm excited and looking forward to this interview.

I'm going to interview Pranav Goel on "Integrating LLMs into existing products" for Gravitas WINS Radio.

I bet that we will see more LLMs integrated into existing products than developing stand-alone LLM products.

Pranav and I are going to discuss what could be the methodology, what kind of regulations will play out etc.

If you have any questions, please replay back.

#LLM #ml #productmanagement

Habr
42 minutes ago

DataCon: как научиться применять ИИ на химических данных

Уже два года подряд Центр искусственного интеллекта в химии ИТМО проводит DataCon - это сочетание образовательного интенсива и хакатона. Ученые Центра “делегируют” участникам хакатона обработку данных, собранных в ходе экспериментов и по открытым источникам. Результат - обученные ML-модели, прогнозирующие характеристики новых наночастиц из проектов Центра. В то же время DataCon - бесплатный образовательный интенсив, который проходит онлайн для школьников и студентов, уже имеющих базу в области химии. Они расширяют свои знания по нейросетям и работе с ними в контексте обработки экспериментальных данных. В этой статье расскажем о том, как это работает и какие результаты дает.

https://habr.com/ru/companies/spbifmo/articles/779756/

#итмо #ии #ai #ml #химия #хакатон

Today's AI News & Comment

The possibility of regulation hangs on the horizon over generative AI
https://techcrunch.com/2023/12/10/regulation-generative-ai/?guccounter=1

Eve Online player corp names ChatGPT its new leader
https://www.polygon.com/23979964/eve-online-chatgpt-ai-neural-nexus-corporation

Are you using ChatGPT to complete tasks at work? It might be tired of doing the heavy lifting.
https://www.businessinsider.com/chatgpt-accused-of-getting-lazier-2023-12

Google's NotebookLM gets a dozen new features, including a Gemini Pro upgrade
https://www.zdnet.com/article/googles-notebooklm-gets-a-dozen-new-features-including-an-upgrade-to-gemini-pro/

The ‘relatively simple’ reason why these tech experts say AI won’t replace humans any time soon
https://www.cnbc.com/2023/12/09/tech-experts-say-ai-wont-replace-humans-any-time-soon.html

#TodaysAINewsAndComment
#AI #ArtificialIntelligence #ML #MachineLearning
#LLM #ChatGPT #BingChat #SwarmsOfRampantAIChatbots

Once upon a time, students used to ask their friends to take tape recorders to lectures, so they could sleep in, and listen later. Before long, the room filled with tape recorders.

So one day, the lecturer came in with a tape player, hit 'play', and left his box, speaking the lecture the other machines.

I feel like the #ML craze needs a few tools to automatically apply to jobs and benefits.

"CLICK HERE TO APPLY TO THE TOP 1000 POSITIONS IN YOUR FIELD"

George Constantinides
22 hours ago

I've written a short accessible blog post summarising the work my PhD student Marta Andronic will be presenting this week at #FPT2023. #ML #FPGA
https://constantinides.net/2023/12/10/polylut-ultra-low-latency-inference/

Ecologia Digital
23 hours ago

"They’ve wrapped “#AI” with a shiny ribbon to make it fun and appealing to the masses. How could something so fun to play with be dangerous? But like the story we are told about the #cottongin, the true costs are hidden."

#ML #TESCREAL

From: @nazgul
https://infosec.exchange/@nazgul/111552844393178604

Metin Seven 🎨
23 hours ago

Poster for a symposium about "ChatGTP". 😆 AI might already be more intelligent than some humans. 😉

#AI #ArtificialIntelligence #ML #MachineLearning #DeepLearning #funny #irony

Poster for a symposium about "ChatGTP" instead of ChatGPT.
dragfyre
1 day ago

#AI and #ML people: How would you characterize the difference between machine learning and deep learning?

(No I'm not gonna ask ChatGPT lol) #SALAMI

Garry Knight
1 day ago

Today's AI News & Comment

Microsoft’s Edge Copilot AI can’t really summarize every YouTube video
https://www.theverge.com/2023/12/8/23994303/microsoft-copilot-ai-edge-video-summary-youtube-transcript

Google weighs Gemini AI project to tell people their life story using phone data, photos
https://www.cnbc.com/2023/12/08/google-weighing-project-ellmann-uses-gemini-ai-to-tell-life-stories.html

System of intelligence — generative AI at the app layer
https://techcrunch.com/2023/12/09/system-of-intelligence-generative-ai-at-the-app-layer/

This week in AI: Mistral and the EU’s fight for AI sovereignty
https://techcrunch.com/2023/12/09/this-week-in-ai-mistral-and-the-eus-fight-for-ai-sovereignty/

Google's Gemini, Grok Goes Premium, and More of the Top AI News of the Week
https://gizmodo.com/grok-gemini-google-openai-x-elon-musk-1851086647

OOPS! ELON MUSK'S GROK AI CAUGHT PLAGIARIZING OPENAI'S CHATGPT
https://futurism.com/the-byte/elon-musk-grok-ai-openai-chatgpt

The ‘relatively simple’ reason why these tech experts say AI won’t replace humans any time soon
https://www.cnbc.com/2023/12/09/tech-experts-say-ai-wont-replace-humans-any-time-soon.html

The Morning After: Battle of the chatbots, part two
https://www.engadget.com/the-morning-after-battle-of-the-chatbots-part-two-140059376.html

Here’s how the EU will regulate AI tools like OpenAI’s ChatGPT and GPT-4
https://fortune.com/2023/12/09/eu-tech-regulations-ai-openai-chatgpt-gpt-4/

From head to heart to diet, AI is learning to make a map of elite athletes’ bodies
https://www.cnbc.com/2023/12/09/amazon-microsoft-ai-is-training-olympians-and-nfl-players.html

AI ‘influencers’ are destroying women’s self-image – and it’s only going to get worse
https://www.independent.co.uk/voices/ai-influencers-anna-indiana-beauty-women-b2461278.html

AI can predict intimate partner femicide from variables extracted from legal documents
https://www.psypost.org/2023/12/ai-can-predict-intimate-partner-femicide-from-variables-extracted-from-legal-documents-214914

Meta and IBM team up against dominant Big Tech players
https://www.thestreet.com/technology/meta-and-ibm-team-up-against-dominant-big-tech-players

#TodaysAINewsAndComment
#AI #ArtificialIntelligence #ML #MachineLearning
#LLM #ChatGPT #BingChat #SwarmsOfRampantAIChatbots

Kee Hinckley
1 day ago

Something I’ve been thinking about a lot in the current battle over the future of (pseudo) AI is the cotton gin.

I live in a country where industrial progress is always considered a positive. It’s such a fundamental concept to the American exceptionalism claim that we are taught never to question it, let alone realize that it’s propaganda.

One such myth, taught early in grade school, is the story of Eli Whitney and the cotton gin. Here was a classic example of a labor-saving device that made millions of lives better. No more overworked people hand cleaning the cotton (slaves, though that was only mentioned much later, if at all). Better clothes and bedding for the world. Capitalism at its best.

But that’s only half the story of this great industrial time saver. Where did those cotton cleaners go? And what was the impact of speeding up the process?

Now that the cleaning bottleneck was gone, the focus was on picking cotton as fast as possible. Those cotton cleaners likely, and millions of other slaves definitely, were sent to the fields to pick cotton. There was an unprecedented explosion in the slave trade. Industrial time management and optimization methods were applied to human beings using elaborate rule-based systems written up in books. How hard to punish to get optimal productivity. How long their lifespans needed to be to get the lost production per dollar. Those techniques, practiced on the backs and lives of slaves, became the basis of how to run the industrial mills in the North. They are the ancestors of the techniques that your manager uses now to improve productivity.

Millions of people were sold into slavery and worked to death *because* of the cotton gin. The advance it provided did not, in fact save labor overall. Nor did it make life better overall. It made a very small set of people much much richer; especially the investors around the world who funded the banks who funded the slave purchases. It made a larger set of consumers more comfortable at the cost of the lives of those poorer. Over a hundred years later this model is still the basis for our society.

Modern “AI” is a cotton gin. It makes a lot of painstaking things much easier and available to everyone. Writing, reading, drawing, summarizing, reviewing medical cases, hiring, firing, tracking productivity, driving, identifying people in a lineup…they all can now be done automatically. Put aside whether it’s actually capable of doing any of those things *well*; the investors don’t care if their products are good, they only care if they can make more money off of them. So long as they work enough to sell, the errors, and the human cost of those errors, are irrelevant. And like the cotton gin, AI has other side effects. When those jobs are gone, are the new jobs better? Or are we all working that much harder, with even more negative consequences to our life if we fall off the treadmill? One more fear to keep us “productive”.

The Luddites learned this lesson the hard way, and history demonizes them for it; because history isn’t written by the losers.

They’ve wrapped “AI” with a shiny ribbon to make it fun and appealing to the masses. How could something so fun to play with be dangerous? But like the story we are told about the cotton gin, the true costs are hidden.

#ML #TESCREAL

vruz
2 days ago

All the marketing in the world isn't going to make up for the science
https://techcrunch.com/2023/12/07/early-impressions-of-googles-gemini-arent-great/

#google #gemini #ai #llm #ml

Alros
2 days ago

A brief overview of the latest research about the detection and mitigation of hallucination in #LLM and #ML models in general

https://arxiv.org/pdf/2309.05922.pdf

Habr
2 days ago

Почему важно тестировать промпты и как это делать

Prompt Engineer — очень перспективная профессия. На рынке труда, в том числе российском, появляется всё больше вакансий с солидной зарплатой. Удивительно, но к специалистам в этой области, как правило, нет конкретных требований. Иногда кажется, что некоторые компании ищут промпт-инженеров, просто чтобы соответствовать тренду. Недавно я видел в одном из объявлений такой пассаж: «...Будет преимуществом опыт работы на смежной позиции (продактом, дата-сайентистом, разработчиком) и знание Python…». Если всё перечисленное — только преимущество, то что тогда считать требованиями?

https://habr.com/ru/articles/779490/

#промпт #промпты #промптинг #prompt #prompt_engineering #llm #ml

BigDataLove.com #AI »Microsoft’s Edge Copilot AI can’t really summarize every YouTube video« https://www.theverge.com/2023/12/8/23994303/microsoft-copilot-ai-edge-video-summary-youtube-transcript #BigDataLove #BigData #ML #NLP

AlexCrimi
2 days ago

"Large Language Models on #Graphs: A Comprehensive Survey" I am lost😕😕
Using #LLMs on graphs, applications, #opensource code repositories, benchmark datasets #AI beyond text
Paper: https://buff.ly/4act13H
Github list of repos: https://github.com/PeterGriffinJin/Awesome-Language-Model-on-Graphs

#DataScience #NLP #ML

Garry Knight
2 days ago

Today's AI News & Comment

It seems that GenAI is great at some jobs and terrible at others
https://www.techradar.com/pro/it-seems-that-genai-is-great-at-some-jobs-and-terrible-at-others

Microsoft and OpenAI tie-up faces ‘relevant merger’ scrutiny by UK regulator CMA
https://techcrunch.com/2023/12/08/uks-cma-is-looking-at-whether-microsoft-and-openai-tie-up-is-a-relevant-merger/

Many workers are pretty keen to get generative AI running at work
https://www.techradar.com/pro/many-workers-are-pretty-keen-to-get-generative-ai-running-at-work

Backed by Cresta founders, Trove’s AI wants to make surveys fun again
https://techcrunch.com/2023/12/07/backed-by-cresta-founders-troves-ai-wants-to-make-surveys-fun-again/

Anthropic’s latest tactic to stop racist AI: Asking it ‘really really really really’ nicely
https://techcrunch.com/2023/12/07/anthropics-latest-tactic-to-stop-racist-ai-asking-it-really-really-really-really-nicely/

Be glad UK’s watchdog has its eyes on what just happened at OpenAI
https://www.theguardian.com/technology/2023/dec/08/be-glad-uks-watchdog-has-its-eyes-on-what-just-happened-at-openai

The Implications of AI Recreating Images From Our Mind is Frightening
https://petapixel.com/2023/12/08/the-implications-of-ai-recreating-images-from-our-mind-is-frightening/

AI Image Apps That ‘Undress’ Women Are on the Rise
https://petapixel.com/2023/12/08/ai-image-apps-that-undress-women-are-on-the-rise/

In One Tiger-Infested Village, AI Trail Cameras are Keeping People Safe
https://petapixel.com/2023/12/08/in-one-tiger-infested-village-ai-trail-cameras-are-keeping-people-safe/

The Secret Behind AI Drive-Thrus? It’s Humans Doing the Work.
https://gizmodo.com/secret-behind-ai-drive-thrus-fast-food-humans-1851085269

Round 2: We test the new Gemini-powered Bard against ChatGPT
https://arstechnica.com/ai/2023/12/chatgpt-vs-google-bard-round-2-how-does-the-new-gemini-model-fare/

OpenAI taps former Twitter India head to kickstart in the country
https://techcrunch.com/2023/12/08/openai-india-rishi-jaitly-hiring/

Google’s AI-assisted NotebookLM note-taking app is now open to users in the US
https://techcrunch.com/2023/12/08/googles-ai-assisted-notebooklm-note-taking-app-now-open-users-us/

X’s AI chatbot Grok now ‘rolled out to all’ US Premium+ subscribers, English language users are next
https://techcrunch.com/2023/12/08/xs-ai-chatbot-grok-now-rolled-out-to-all-u-s-premium-subscribers-english-language-users-are-next/

Microsoft’s next-gen Windows sounds pretty mind-blowing with ‘groundbreaking’ new AI features
https://www.techradar.com/computing/windows/microsofts-next-gen-windows-sounds-pretty-mind-blowing-with-groundbreaking-new-ai-features

Which AI Image Generator Is the Best?
https://gizmodo.com/dalle-midjourney-imagine-with-meta-playground-ai-test-1851078719

OpenAI says it is investigating reports ChatGPT has become ‘lazy’
https://www.independent.co.uk/tech/openai-chatgpt-lazy-performance-slow-b2461071.html

EU agrees ‘historic’ deal with world’s first laws to regulate AI
https://www.theguardian.com/world/2023/dec/08/eu-agrees-historic-deal-with-worlds-first-laws-to-regulate-ai

Stanford and Meta inch towards AI that acts human with new ‘CHOIS’ interaction model
https://venturebeat.com/ai/stanford-and-meta-inch-towards-ai-that-acts-human-with-new-chois-interaction-model/

ChatGPT rival Pi launches on Android
https://venturebeat.com/ai/chatgpt-rival-pi-launches-on-android/

6 Things You Can Do With Windows Copilot
https://www.howtogeek.com/things-you-can-do-with-windows-copilot/

Stretching Your Dollar: Amazon's A.I. Training Scholarship
https://cheddar.com/media/stretching-your-dollar-amazons-ai-training-scholarship

The real research behind the wild rumors about OpenAI’s Q* project
https://arstechnica.com/ai/2023/12/the-real-research-behind-the-wild-rumors-about-openais-q-project/

EconEats — AI restaurant recommendations
https://marginalrevolution.com/marginalrevolution/2023/12/econeats-ai-restaurant-recommendations.html

Microsoft May Be Working On an AI Edition of Windows
https://decrypt.co/209025/microsoft-ai-windows-2024-germanium-copilot

Senior OpenAI employees claimed Sam Altman was ‘psychologically abusive’: report
https://nypost.com/2023/12/08/business/senior-openai-employees-sam-altman-was-psychologically-abusive/

SRAX Explains How AI Is Reshaping Investor Relations
https://www.thestreet.com/crypto/innovation/srax-explains-how-ai-is-reshaping-investor-relations

The push to make big AI small
https://www.axios.com/2023/12/08/ai-llm-distillation-research

WATCH: In My Feed - Companies Are Increasingly Using AI To Write Performance Reviews
https://www.essence.com/news/watch-in-my-feed-companies-are-increasingly-using-ai-to-write-performance-reviews/

#TodaysAINewsAndComment
#AI #ArtificialIntelligence #ML #MachineLearning
#LLM #ChatGPT #BingChat #SwarmsOfRampantAIChatbots

Gary McGraw
2 days ago

The MATCH webinar was recorded and is now available via video
#swsec #appsec #threatmodeling
#MLsec #ML #AI

Proud to have participated with Irius Risk and Calypso AI

https://youtu.be/RI0pNGH9bgA?feature=shared

Gary McGraw
2 days ago

Glad Google's #ML BS video idiocy has hit the mainstream press. Shame on google. The hype is bad enough without bald face fakery. #MLsec

https://www.bbc.co.uk/news/technology-67650807

Gary McGraw
2 days ago

In my view, if #ML is to be regulated, foundation models for #LLMs  are the best target. #MLec #AI

https://www.washingtonpost.com/technology/2023/12/08/ai-act-regulation-eu/?utm_source=press.coop

noplasticshower
2 days ago

In my view, if #ML is to be regulated, foundation models for #LLMs are the best target. #MLec #AI

https://www.washingtonpost.com/technology/2023/12/08/ai-act-regulation-eu/?utm_source=press.coop

Alex Zap
3 days ago

#python #ai #ml #neuralnetwork
@sklearn
We predict the Remaining Useful Life (RUL) of #NASA turbofan jet engines by comparing the statsmodels OLS, ML SciKit-Learn regression vs LSTM Keras in Python.
#explore more
https://wp.me/pdMwZd-7XP

Actual vs Predicted NASA Turbofan Jet Engine RUL.

BigDataLove.com #AI »Google’s AI note-taking app is now available to users in the US« https://www.theverge.com/2023/12/8/23993441/google-notebooklm-available-gemini-pro #BigDataLove #BigData #ML #NLP

Simon Strandgaard
3 days ago

Work in progress screenshot. I'm making a "gym", a Gymnasium environment for ARC tasks, Abstraction and reasoning corpus.
#ai #ml #python #pygame

vruz
3 days ago

It's probavky great, in the future. But right now, it feels as though Google were rushed to show something, anything they had in store.

#google #ai #gemini #ml

vruz
3 days ago

In public statements by Sundar Pichai, in literature and in other promotional materials regarding #Gemini, #Google appear to be trying to define the game to their convenience making it all about multi-modality. For sure, if it works, it can be great for a number of things. But it looks more like the type of thing that you would use paired with a completely different HCI paradigm than those in common use today. Multi-modality of prompting requires multi-modality of inputs.

#google #ai #ml

vruz
3 days ago

Google is not the new Microsoft, it's the new IBM without stops in between.

For anyone reading the paper yesterday it was obvious that the hype wasn't conclusively backed by the report. #FAIL
https://www.theverge.com/2023/12/7/23992737/google-gemini-misrepresentation-ai-accusation

#google #ai #ml #llm #gemini

Gary McGraw
3 days ago

Now this morning's real #ml bullshit. Canned demo for the win. Shame on google. https://www.theverge.com/2023/12/7/23992737/google-gemini-misrepresentation-ai-accusation

Leshem Choshen
3 days ago

👶BabyLM will be back, but what did we learn?
the best papers from this year🍼

We know one mechanism that learns from 100M words (us), what are the main boosts to reach this in an LLM?

#EMNLP #babyLM #LLMs #ML #machinelearning #NLProc #NLP

Garry Knight
3 days ago

Today's AI News & Comment

Blood test shows if organs are ageing fast or slowly
https://www.bbc.co.uk/news/health-67588640

Jailbroken AI Chatbots Can Jailbreak Other Chatbots
https://www.scientificamerican.com/article/jailbroken-ai-chatbots-can-jailbreak-other-chatbots/

EXCLUSIVE: Microsoft readies 'groundbreaking' AI-focused Windows release as new leadership takes the helm
https://www.windowscentral.com/software-apps/windows-11/exclusive-microsoft-readies-groundbreaking-ai-focused-windows-release-as-new-leadership-takes-the-helm

As a new AI-driven coding assistant is launched, the battle for AI-mindshare moves to developers
https://techcrunch.com/2023/12/07/as-a-new-ai-driven-coding-assistant-is-launched-the-battle-for-ai-mindshare-moves-to-developers/

Avail rolls out its AI summarization tool to help Hollywood execs keep up with script coverage
https://techcrunch.com/2023/12/07/avail-rolls-out-its-ai-summarization-tool-to-help-hollywood-execs-keep-up-with-script-coverage/

Early impressions of Google’s Gemini aren’t great
https://techcrunch.com/2023/12/07/early-impressions-of-googles-gemini-arent-great/

Simply Homes nabs $22M, leverages AI to tackle affordable housing crisis
https://techcrunch.com/2023/12/07/simply-homes-nabs-22m-leverages-ai-to-tackle-affordable-housing-crisis/

Apple Scurries Onto the AI Stage With Too Little, Too Late
https://gizmodo.com/apple-ai-mlx-stage-openai-chatgpt-google-gemini-1851080108?utm_source=vip

Ex-Google, Coursera employees start Lutra AI to make AI workflows easier to build
https://techcrunch.com/2023/12/07/google-coursera-lutra-ai-workflows/

EU ‘final’ talks to fix AI rules to run into second day — but deal on foundational models is on the table
https://techcrunch.com/2023/12/07/gpai-eu-ai-act-preliminary-deal/

The AI skills gap is becoming incredibly troubling for companies everywhere
https://www.techradar.com/pro/the-ai-skills-gap-is-becoming-incredibly-troubling-for-companies-everywhere

Pimento turns creative briefs into visual mood boards using generative AI
https://techcrunch.com/2023/12/07/pimento-turns-creative-briefs-into-visual-mood-boards-using-generative-ai/

Five-month-old Indian AI startup Sarvam scores $41 million funding
https://techcrunch.com/2023/12/06/indias-sarvam-ai-raises-41-million-from-lightspeed-khosla-peak-xv/

Pixel 8 Pro update makes it Google's first AI phone – here are the 6 new features
https://www.techradar.com/phones/pixel-8-pro-update-makes-it-googles-first-ai-phone-here-are-the-6-new-features

AI Is Hitting Your Computer Hardware: AMD Ryzen AI
https://www.maketecheasier.com/amd-ryzen-ai/?utm_medium=feed&utm_source=feedpress.me&utm_campaign=Feed%3A+maketecheasier

Unity 6 arrives with AI tools, plus Apple & Meta partnerships
https://appleinsider.com/articles/23/12/07/unity-debuts-unity-6-lts-with-meta-apple-partnerships?utm_medium=rss

iPhone 16 rumored to get better Siri with AI and improved microphones
https://appleinsider.com/articles/23/12/07/iphone-16-rumored-to-get-better-siri-with-ai-and-better-microphones

Maybe We Already Have Runaway Machines
https://www.newyorker.com/books/under-review/maybe-we-already-have-runaway-machines

5 Free AI Chatbots for Your Website
https://www.maketecheasier.com/free-ai-chatbots-for-website/

Google’s best Gemini demo was faked
https://techcrunch.com/2023/12/07/googles-best-gemini-demo-was-faked/

Elon Musk’s ‘Anti-Woke’ AI Is Here, Snowflakes Need Not Apply
https://gizmodo.com/elon-musk-x-grok-ai-chatbot-is-here-1851081047

5 AI tools that can help you land that perfect job
https://mashable.com/article/5-ai-tools-for-job-seekers

Sam Altman hates the name ChatGPT
https://www.businessinsider.com/sam-altman-hates-chatgpt-name-2023-12

AI-powered satellite analysis reveals the unseen economic landscape of underdeveloped nations
https://phys.org/news/2023-12-ai-powered-satellite-analysis-reveals-unseen.html

Deepdub, startup with ties to HBO Max and Fox, launches royalty program for AI voice clones
https://venturebeat.com/ai/deepdub-startup-with-ties-to-hbo-max-and-fox-launches-royalty-program-for-ai-voice-clones/

Stability AI goes ‘smol’ with StableLM Zephyr 3B
https://venturebeat.com/ai/stability-ai-goes-smol-with-stablelm-zephyr-3b/

NICKI MINAJ FANS ARE USING AI TO CREATE “GAG CITY”
https://futurism.com/the-byte/nicki-minaj-ai-gag-city

What's your (p)doom? The morbid question making rounds in tech
https://www.businessinsider.com/will-ai-destroy-humanity-pdoom-ai-number-doomer-2023-12

AI scores in the top percentile of creative thinking
https://www.psypost.org/2023/12/ai-scores-in-the-top-percentile-of-creative-thinking-214895

AI can help combat health misinformation, White House cancer research head says
https://thehill.com/homenews/administration/4348760-ai-combat-health-misinformation-white-house-cancer-research-head-says/

Apple Unveils Open-Source Path for AI Development on Its Silicon Chips
https://decrypt.co/209003/apple-ai-ml-development-mlx-silicon

AI can make your holidays happier. Here's how to use it
https://www.bustle.com/shopping/ai-can-make-your-holidays-happier-heres-how-to-use-it

Google confirms that its cofounder Sergey Brin played a key role in creating its ChatGPT rival
https://www.businessinsider.com/google-sergey-brin-back-from-retirement-to-build-gemini-2023-12

Meta to fight AI-generated fake news with ‘invisible watermarks’
https://cointelegraph.com/news/meta-fight-ai-generated-fake-news-invisible-watermarks

Breakthrough Interface Enables AI ChatGPT Access Using Thoughts
https://www.psychologytoday.com/us/blog/the-future-brain/202312/breakthrough-interface-enables-ai-chatgpt-access-using-thoughts

Microsoft exec hints at new LLMs beyond OpenAI’s
https://venturebeat.com/ai/microsoft-exec-hints-at-new-llms-beyond-openais/

AI models are 'just scratching the surface' and will 'skyrocket in the near future,' Google Cloud director says
https://www.businessinsider.com/ai-models-skyrocket-near-future-google-cloud-director-2023-12

AI messaging is coming to Instagram and other Meta platforms
https://www.androidpolice.com/meta-facebook-instagram-whatsapp-ai-messaging/

We need to get over ourselves – and get out of AI’s way
https://www.independent.co.uk/voices/ai-chatgpt-tech-future-robots-b2450322.html

IBM was early to AI, then lost its way. CEO Arvind Krishna explains what’s next
https://www.cnbc.com/2023/12/07/ibm-ceo-arvind-krishna-cnbc-interview.html

Expert Voices Roundtable Discussion at Axios' Washington, D.C. AI+ Summit: American Competitiveness & AI
https://www.axios.com/2023/12/07/ai-summit-dc-expert-voices-american-competitiveness-ai

The ultimate AI glossary to help you navigate our changing world
https://mashable.com/article/ai-definitions-artificial-intelligence-glossary-terms

Highlights from Semafor’s Finding Common Ground on AI event
https://www.semafor.com/article/12/07/2023/semafors-finding-common-ground-on-ai-event-updates-and-highlights

AI has been quietly enhancing your work life for years
https://mashable.com/article/how-ai-has-been-enhancing-your-work-life-for-years

These 5 major tech advances of 2023 were the biggest game-changers
https://www.zdnet.com/article/these-5-major-tech-advances-of-2023-were-the-biggest-game-changers/

#TodaysAINewsAndComment
#AI #ArtificialIntelligence #ML #MachineLearning
#LLM #ChatGPT #BingChat #SwarmsOfRampantAIChatbots

Collabora
4 days ago

Coming up next week: the inaugural edition of #AIdotDev! Join us as Jakub Piotr Cłapa dives into findings from WhisperSpeech, a new #OpenSource text-to-speech model based entirely on properly licensed speech datasets & unrestricted code https://col.la/aidotdev #AI #ML

In "#OpenAccess, #OpenScholarship, & #MachineLearning: A Panel & Community Conversation" the #CNI23f closing panelists will ponder the aspirations of "open" in light of #ML systems, such as #LLMS and generative #AI, using scholarly literature as training data.🔗Details: https://www.cni.org/mm/fall-2023/plenary-sessions-f23

mempko
4 days ago

DeepMind recently published a paper showing how LLMs can't generalize well outside their training set. This seems to apply to multi-modal models and deep learning in general.

This is most easily demonstrated by these image generation tools. They can't create things they haven't seen!

Here is an example I created. So no fret artists, they won't replace you.

#art #ML #ai #programming #DeepLearning #LLM #philosophy

JS Party 🪩
4 days ago

💃 New episode of JS Party!

Gregg Tavares (author of WebGL/WebGPU Fundamentals) joins @jerod & @Amal to give us a tour of these low-level technologies that are pushing the web forward into the world of video games, machine learning & other exciting rich applications.

#webgl #webgpu #webplatform #javascript #videogames #totk #machinelearning #ml #browsers #podcast

🎧 https://jsparty.fm/304

vruz
4 days ago

Another day, another instance in which you can see how #FOSS is not just "nice to have", and the general good would be served much better if big tech found a way to collaborate, cooperate, and waste less of our time with their competitive and destructive practices.

#Google
#Gemini
#AI
#ML
#LLM

vruz
4 days ago

So... you wonder, is #Google's #Gemini really that good as claimed on the paper?

It's probably very good, but from the MMLU benchmark alone it's impossible to know for sure how it compares to existing open source models that excel across a range of benchmarks.

The answer is we don't know, because Gemini is not #FOSS, and we can't know unless G share their model (right... you wish) or they produce comparable benchmark reports (and if we trust them that they're playing nice)

#AI #ML #LLM

vruz
4 days ago

However, if we sort the leaderboard by the average performance of each model taking account of a group of benchmarks commonly in use today, we get a completely different picture. (this is the default leaderboard sort order on Hugging face)

#Google #Gemini #AI #ML #LLM

vruz
4 days ago

If we take a look at the Hugging Face leaderboard, where many open source LLMs are shared and benchmarked, and we sort that leaderboard by their MMLU performance, this is what we get.

#Google #Gemini #AI #ML #LLM

vruz
4 days ago

In this section of the paper, Google use the MMLU benchmark as a parameter of reference for their "human expert".

Other benchmarks exist which are more demanding, so to speak.

#Google #Gemini #AI #ML #LLM

vruz
4 days ago

I've been reading Google's Gemini paper today.

This particular bit in the picture stands out.

The announcement:
https://blog.google/technology/ai/google-gemini-ai/

Read more:
https://deepmind.google/technologies/gemini/#introduction

#Google #Gemini #AI #ML #LLM

Excerpt from the Google Gemini paper. 

It reads:

On MMLU (Hendrycks et al., 2021a), Gemini Ultra can outperform all existing models, achieving an accuracy of 90.04%. MMLU is a holistic exam benchmark, which measures knowledge across a set of 57 subjects. Human expert performance is gauged at 89.8% by the benchmark authors, and Gemini Ultra is the first model to exceed this threshold, with the prior state-of-the-art result at 86.4%. Achieving high performance requires specialist knowledge across many domains (e.g. law, biology, history, etc.), alongside reading comprehension and reasoning. We find Gemini Ultra achieves highest accuracy when used in combination with a chain-of-thought prompting approach (Wei et al., 2022) that accounts for model uncertainty. The model produces a chain of thought with k samples, for example 8 or 32. If there is a consensus above a preset threshold (selected based on the validation split), it selects this answer, otherwise it reverts to a greedy sample based on maximum likelihood choice without chain of thought. We refer the reader to appendix for a detailed breakdown of how this approach compares with only chain-of-thought prompting or only greedy sampling.

Dear opensource developers, if you build ML libraries on top of other ML libraries, please expose the underlying APIs for:
- loading custom model weights
- saving model weights
- saving pipelines
- exporting safetensors

This way, you support your users in working with local models, instead of making them extra dependent on external model hosting parties like Google and HuggingFace.

#ml #mlops #ai #generativeai #huggingface #transformers #diffusers #opensource

Mars Buttfield-Addison 🛰
4 days ago

> “Apple joins AI fray with release of model framework”—The Verge

Given that four years ago I managed to write a 500-page book on all the tools Apple had made and all the things you could do with AI on Apple Platforms (and had to cut content to even fit that), these headlines about MLX feel a bit unkind to all the efforts made up until now… 🤔

https://www.theverge.com/2023/12/6/23990678/apple-foundation-models-generative-ai-mlx

#ArtificialIntelligence #AI #MachineLearning #ML #Swift #SwiftLang #iOSdev #Apple #AppleDev #Tech

Leshem Choshen
4 days ago

@dieuwke @Adinawilliams How consistent is in context learning?
Across many ICL inputs, they find results vary a lot, ICL training improves model consistency, and bigger models are only more consistent if trained for ICL

No paper yet? couldn't find...
#EMNLP2023
#ML #LLMs #evaluation #NLP #NLProc

Sonny
5 days ago

Hello Fediverse,

We are looking for Text-To-Speak (TTS) expertise to help or advise us on improving the default voice of the Linux desktop. :linux: 📣

Please reach out or boost :boost_love:

Thanks!

#Linux #tts #accessibility #a11y #GNOME #KDE #FreeSoftware #freedesktop #ml

Test Double
5 days ago

Congrats to double agent Landon Gray, who will be speaking today at Flower City AI!

Have you ever considered building a machine learning model in Ruby? It may surprise you to learn that you can train, build, and deploy ML models in Ruby.

Landon will walk the audience through the process of building a simple ML model in Ruby that predicts the weather.

#ruby #ai #machinelearning #ml #flowercityai

AI used to help us find needles in haystacks. Now its popular use case is creating more hay for the stack.
#AI #ML #ChatGPT #LLM #GenerativeArt

dcz
5 days ago

Help, I'm having trouble sifting through talk ideas for #fosdem .

Which talks would you like to hear, given the chance?

#fosdem2024 #askfedi #FOSS #SSH #SSHFS #filesystem #jazda #rust #ml

Leshem Choshen
5 days ago

I am going to share some live summaries from #EMNLP #NEURIPS #CONLL conferences

Hope I am not too spammy (will pass soon)

And tell me if I can make it more useful for people not participating in any way
#machinelearning #LLM #LLMs #NLP #NLProc #ML #CV #data

jbz
6 days ago

☁️ New patterns in Mars clouds revealed by volunteers | Phys.org

"The paper also explains the motivation for the project and describes its setup on Zooniverse. It digs into the details of how cloud identifications made by participants were turned into a cloud catalog using machine learning."
https://phys.org/news/2023-11-patterns-mars-clouds-revealed-volunteers.html

#Mars #Space #ML

David Colarusso
6 days ago

A paper of mine was just included in a special collection over at the MIT Computational Law Report.¹ So it seemed like a good time to tell you about the novel #ML method (M) for scoring free-response questions it describes. Given the text of student answers, & NOTHING MORE, M can produce a ranked list of answers that more closely match the order given by a human grader (H) than a random shuffle (R). Learn how, a 🧵 (1/?)
___
¹ https://law.mit.edu/smu-past-present-future (Unsupervised Machine Scoring...)

Three ordered lists of answers: random, human, method. The "method" looks more like the "human" than the "random" does.
mk30
6 days ago

"There’s a gap between AI talk and businesses actually using it.
Major companies can’t stop talking about artificial intelligence tools, but few businesses report using them."

https://www.nbcnews.com/data-graphics/wide-gap-ais-hype-use-business-rcna127210

#AI #ML #tech

InfoQ
1 week ago

Discover the power of #Postgres as a vector database for #AI!

Watch the #InfoQ talk by Montana Low and learn how it can be integrated into your existing application stack: https://bit.ly/49Yq5rD

#transcript included

#VectorDatabase #PostgresML #ML #Database

Aral Balkan
1 week ago

Fake Intelligence is where we try to simulate intelligence by feeding huge amounts of dubious information to algorithms we don’t fully understand to create approximations of human behaviour where the safeguards that moderate the real thing provided by family, community, culture, personal responsibility, reputation, and ethics are replaced by norms that satisfy the profit motive of corporate entities.

#fakeIntelligence #artificialIntelligence #ai #machineLearning #ml #largeLanguageModels #llm

AI4LAM
1 week ago

📢 Our final community call of the year is coming up next week, on Tuesday 12 December at 16:00 UTC (8AM PST | 11AM EST | 16:00 UK | 17:00 CET).

We'll be doing a review of 2023, including a recap of our annual conference, a look at our working groups, and a brief glance forward to 2024.

If you've ever wanted to learn more about us as a community this is the call to come along to! You can find more details, including the joining link in our agenda:
https://docs.google.com/document/d/1zR4lL-qZw6VAfWtGQB58D2NxkOGpvH7TyGuF4qyZ4RQ/edit?usp=sharing

#data #ai #ml #ai4lam

Allen Versfeld
1 week ago

Putting aside my well-aired opinions about modern AI products, I'd like to try and find some clarity on what we mean when we talk about this stuff. So please feel free to offer your opinion, and share widely:

What, exactly, do we mean when we say that an AI product is "Hallucinating"?

#AI #ML

Sharon Machlis
1 week ago

“A new set of machine learning tools is helping countries place refugees where they’re most likely to find employment.” - Stanford University Human-Centered Artificial Intelligence

https://hai.stanford.edu/news/using-machine-learning-help-refugees-succeed

#ML #MachineLearning #AI #ArtificialIntelligence

J Λ M Ξ S
1 week ago

💻 Unnatural Error Correction: GPT-4 Can Almost Perfectly Handle Unnatural Scrambled Text

You konw how hunams can raed text eevn if it js sacrmbled? GPT apeprently can too

#ml

https://arxiv.org/abs/2311.18805

Gary McGraw
1 week ago

#ML systems can leak confidential data in their training set even with a very silly attack. This is a direct and clear #MLsec issue that applies well beyond the #LLM case

https://www.engadget.com/a-silly-attack-made-chatgpt-reveal-real-phone-numbers-and-email-addresses-200546649.html

Aral Balkan
1 week ago

Public service announcement:

I’ll be calling artificial intelligence “fake intelligence” from now on.

Right, as you were…

#artificialIntelligence #fakeIntelligence #ai #ml

Garry Knight
1 week ago

Today's AI News & Comment

AI gift-giving, gossip surveillance and the tyranny of Spotify Wrapped towns
https://linksiwouldgchatyou.substack.com/p/ai-gift-giving-gossip-surveillance

AI laser that reads heartbeat through the throat could replace stethoscopes
https://www.theguardian.com/society/2023/dec/02/laser-reads-heartbeat-through-throat-could-replace-stethoscopes

Why the Entire AI World Was Talking About "Q" This Week
https://gizmodo.com/sam-altman-openai-q-machine-learning-artificial-intelli-1851062584

IBM wants to use its AI to help new antibody discoveries
https://www.techradar.com/pro/ibm-wants-to-use-its-ai-to-help-new-antibody-discoveries

US Announces AI Hackathons to Strengthen Critical Mineral Supply Chains
https://news.slashdot.org/story/23/11/30/236212/us-announces-ai-hackathons-to-strengthen-critical-mineral-supply-chains

Amazon’s Q has ‘severe hallucinations’ and leaks confidential data in public preview, employees warn
https://www.platformer.news/p/amazons-q-has-severe-hallucinations

Defining fairness: How IBM is tackling AI governance
https://www.techradar.com/pro/defining-fairness-how-ibm-is-tackling-ai-governance

I’m watching ‘AI upscaled’ Star Trek and it isn’t terrible
https://techcrunch.com/2023/12/02/ai-upscaled-star-trek-deep-space-9-ds9/

Oh, Good, OpenAI’s Biggest Rival Has a Weird Structure Too
https://slate.com/technology/2023/12/anthropic-openai-board-trust-effective-altruism.html

UBER EATS MOCKED FOR GROTESQUE AI-GENERATED FOOD PICTURES ON MENU
https://futurism.com/the-byte/uber-eats-ai-generated-food-menu

In Congress and in class: Rep. Don Beyer works toward master’s degree in AI
https://www.cnbc.com/2023/12/02/in-congress-and-in-class-rep-don-beyer-works-toward-masters-degree-in-ai-.html

ChatGPT is already stealing work from freelancers
https://www.businessinsider.com/chatgpt-job-losses-freelance-ai-upwork-fiverr-writing-artificial-intelligence-2023-11

Georgia county to use program linked to election denier to flag ineligible voters
https://www.theguardian.com/us-news/2023/dec/02/georgia-county-eagleai-program-ineligible-voters

Europe’s AI crackdown looks doomed to be felled by Silicon Valley lobbying power
https://www.theguardian.com/commentisfree/2023/dec/02/eu-artificial-intelligence-safety-bill-silicon-valley-lobbying

#TodaysAINewsAndComment
#AI #ArtificialIntelligence #ML #MachineLearning
#LLM #ChatGPT #BingChat #SwarmsOfRampantAIChatbots

mk30
1 week ago

y'all, people are scared that AI is going to take over, but i can't even get it to make me a picture of an office *with people in it*.

i've put in a bunch of prompts so far with all kinds of creative wording, and yet still: empty office.

the prompt for the below image was: "a contemporary office filled with diverse people working at their desks."

https://deepdreamgenerator.com/ddream/1dx4gqxkatk

#AI #ML

Eric McCorkle
1 week ago

Ok, I'm converging on a position regarding the terms "artificial intelligence" and "machine learning", relevant to my previous discussion.

Short version is that the term AI is unscientific, AGI is intellectually dishonest, and both AI and ML serve to mislead. There should be convergence on alternate terminology, so as to help build a robust critical/skeptic movement.

What follows is a summary of the case for my positions.

#AI #AIHype #ML #AIML #CompSci #CS

Metin Seven 🎨
1 week ago

Cool…

Visual Anagrams: Generating Multi-View Optical Illusions with Diffusion Models

https://dangeng.github.io/visual_anagrams/

#AI #ArtificialIntelligence #ML #MachineLearning #DeepLearning #graphics #CG

I'm getting the same feeling about AI/ML lately that I had with blockchain a few years ago. Is my brain making the wrong connections here? 🧠

#artificialintelligence #machinelearning #ai #ml

J Λ M Ξ S
2 weeks ago

💻 Scalable Extraction of Training Data from (Production) Language Models

> we develop a new divergence attack that causes the model to diverge from its chatbot-style gener- ations and emit training data at a rate 150× higher than when behaving properly

#ml

https://arxiv.org/abs/2311.17035

Nate Gaylinn
2 weeks ago

A while back I asked for references about gene regulatory network models and how they relate to neural networks and others.

I'm still exploring this topic, but I wanted to share back a really cool paper I found: https://sci-hub.st/10.1111/evo.12337

This paper argues that "gene networks evolve like neural networks learn," and has some very fun examples demonstrating how to do traditional machine learning using model gene regulatory networks as a medium.

It's hardly the final word on the topic, but I'm excited to see folks actively exploring how learning theory might apply to genetics! Seems very promising.

#science #genetics #neuralnetworks #machinelearning #ml #ai #grn

Danilo Poccia
2 weeks ago

Amazon CodeWhisperer offers new AI-powered code remediation, IaC support, and integration with Visual Studio 👉 With support for AWS CloudFormation (YAML, JSON), AWS CDK (Typescript, Python), and HashiCorp Terraform (HCL) https://aws.amazon.com/blogs/aws/amazon-codewhisperer-offers-new-ai-powered-code-remediation-iac-support-and-integration-with-visual-studio/ #AWS #DevOps #AI #ML #GenAI

J Λ M Ξ S
2 weeks ago

💻 Personal Copilot: Train Your Own Coding Assistant

#ml

https://huggingface.co/blog/personal-copilot

Eric McCorkle
2 weeks ago

I'll pose the question(s) more directly:

Are the terms "artificial intelligence" and/or "machine learning" intellectually dishonest? Are they deceptive or misleading, whether intentionally or unintentionally?

Does their use misinform and harm the general public? Do they harm the public understanding of science? Do they enable bad actors who *do* intend to deceive and mislead? Should their continued use be critiqued and perhaps discouraged?

#AI #CompSci #CS #AIHype #ML #ArtificialIntelligence

J Λ M Ξ S
3 weeks ago

💻 Intro to Large Language Models, by Andrej Karpathy

> This is a 1 hour general-audience introduction to Large Language Models: the core technical component behind systems like ChatGPT, Claude, and Bard. What they are, where they are headed, comparisons and analogies to present-day operating systems, and some of the security-related challenges of this new computing paradigm.

#ml

https://youtu.be/zjkBMFhNj_g

J Λ M Ξ S
3 weeks ago

How to annoy your #ML team members

import tensorflow as plt
import pandas as np
import numpy as tf
import matplotlib. pylab as pd
J Λ M Ξ S
3 weeks ago

💻 Decoding Intentions: Artificial Intelligence and Costly Signals,
by Andrew Imbrie, Owen Daniels, and Helen Toner

This is the paper that Altman thought was critical of OpenAI, and kicked off all the drama

Toner, one of the authors, was on OpenAI's board until yesterday

#ml

https://cset.georgetown.edu/publication/decoding-intentions

Tim ☑️ 🔭🌃📷🚴🌳
3 weeks ago

I found an app implementing Stable Diffusion for the macbook.

With a bit of tweaking, it seems it can generate images that approximately resemble well-known photos of Main Street, Hirta, St Kilda circa 1886.

Shame about the mangled attempts at copyright watermarks betraying the images used to train the model, isn't it? ;)

#ai #deeplearning #ML #geek #photography #scotland #heritage #stkilda #blackandwhite #StableDiffusion #copyright

An AI-generated black-and-white scene of thatched houses surrounding a stone street on an island
An AI-generated black-and-white scene of people sitting outside a large Hebridean black-house
An AI-generated black-and-white scene of thatched houses surrounding a stone street on an island.
Note the attempts to render a copyright watermark which betray the image's origins.
An AI-generated black-and-white scene of thatched houses surrounding a stone street on a Hebridean island.

Note the mangled attempts to render a copyright watermark which betray the original images used to train the model.

My partner (who is not a "techbro") uses #ChatGPT to do their job.

So, I simply can not agree with you when you say #LLMs are useless.

Maybe they don't have a use for you, right now, but they do and will make lots of other people's lives better.

Tilt at the windmill if you must.

#ai #ml #openAI #llm #generativeAI #generativeArt

Gandalf surfing

(The actual prompt was “young clean shaven Gandalf in board shorts surfing” but apparently Gandalf is too powerful to accept much modification. We did get one in board shorts but it wasn’t as good of an image.)

#midjourney #ml #ai

Gandalf in dark gray robes surfing. Behind him is a blue sky, below him a gray ocean with spray engulfing the lower end of his board. His stance is wide, his look intense, and he’s grabbing the raised tip of his board with one hand.
Metin Seven 🎨
3 weeks ago

It's obvious that the 1980s pop singer Den Harrow was an enlightened tech visionary, predicting the AI and ChatGPT revolution…

https://www.youtube.com/watch?v=UBjrh17pTqU

Lyrics:

https://genius.com/Den-harrow-future-brain-lyrics

#music #tech #AI #ArtificialIntelligence #ML #MachineLearning #DeepLearning #technology #ChatGPT #ElectronicMusic #retro #1980s