Israel uses AI in fight against Hamas

After the horrific attacks on Israeli civilians this past weekend, Hamas launched a follow-up barrage of thousands of rockets. More than 90% were blown up in mid-air by Israel’s AI-powered “Iron Dome” missile defense system. The Israel Defense Forces (IDF) are using AI for a number of critical tasks, including identifying rocket launchpads from drone footage and satellite images, targeting terrorist leaders using facial recognition, and organizing and deploying swarms of attack drones.

Israel uses AI in the fight against Hamas

ChatGPT’s mobile app is raking in the $$$

ChatGPT’s mobile app on iPhone and Android generated $4.6 million in revenues in September. The app has reached over 15 million downloads in the 4 months since its launch.

ChatGPT mobile rakes in $$$

Bing AI comes to Chrome, challenging Bard on its home turf

Microsoft had hoped that its nifty Bing AI chatbot/search engine would lure users to switch to Microsoft’s Edge browser. That hasn’t happened, so Microsoft has now pivoted to a Bing-on-every-browser strategy. This includes the world’s leading browser Chrome, where Bing will compete head to head with Google’s Bard on Google’s home turf.

Bing challenges Bard on Chrome

AMD acquires open source AI pioneer Nod.ai 

Industry-leading GPU chipmaker NVidia has been on an acquisition spree, putting its fingers in every part of the AI pie. Now rival AMD tells Nvidia, “Hold my beer,” as it acquires pioneering open source AI software maker Nod.ai. Nod specializes in optimizing AI software for high performance hardware - like AMD’s and Nvidia’s.

AMD acquires open source AI pioneer Nod ai

The 7 fundamental use cases for GPT-4V

Twitter/X has been awash with videos showing use cases for OpenAI’s new multimodal ChatGPT-4V. Now @GregKamradt has developed a framework for classifying all of these use cases into 7 broad categories: Describe, Interpret, Recommend, Convert, Extract, Assist, or Evaluate.

GPT-4V’s 7 fundamental use cases

Open source LLaVA competes with OpenAI’s multimodal GPT-4V

OpenAI’s latest release, the multimodal GPT-4V, has been all the rage in the 2 weeks since it was released.

Now comes an open source competitor, LLaVA 1.5, and early reports indicate that it’s pretty freaking good. Git over to GitHub and check it out.

First impressions of LLaVA 1.5

LLaVA comparison with GPT-4V

ElevenLabs will dub your voice into 20 different languages

Year-old voice cloning startup ElevenLabs, founded by ex-employees of Google and Palantir, have just announced AI Dubbing, which will dub your voice into 20 different languages, while retaining your individual voice characteristics. So now you can distribute your podcast in Hindi and Arabic, as well as English.

ElevenLabs dubs your voice into 20 different languages

AI could soon use as much electricity as a small country

A researcher at the Amsterdam School of Business and Economics has published a paper projecting that by 2027, AI could use approximately 0.5% of the world’s electricity, similar to the total usage of small nations such as Argentina or Sweden.

NYT: AI could soon use as much electricity as a small country

Peer-reviewed source article

Disney’s Loki poster generates anti-AI backlash

Disney wanted to create buzz around the upcoming second season of Loki. Be careful what you wish for. Sharp-eyed readers detected telltale traces of AI generation in the striking image on the promo poster, and graphic designers cried foul.

Disney faces anti-AI backlash over Loki poster

Google uses AI to make traffic lights less annoying

More than a dozen cities across the globe are engaged in a pilot program with Google, using information distilled from traffic data gleaned by its popular Maps app, in order to develop location-specific timing sequences for individual traffic signals. The goal is to reduce idling time, decrease wasted fuel, and thereby reduce emissions. Preliminary results from Seattle, one of the participating cities, indicates that the AI-optimized timing sequences saved 10% of emissions from 30 million cars per month.

Google uses AI to make traffic lights less annoying

rabbit raises $20 million for new AI-powered operating system

Premier VC firm Khosla Ventures led a $20 million round of funding for rabbit, a startup focused on building an operating system for the Age of AI, featuring a natural language interface and a focus on DWIW (“Do what I want”) rather than the prevailing OS standard of WYSIWYG (“What you see is what you get”).

AI startup rabbit snags $20 million for AI-powered OS

AI makes everyone a coder

LLMs are unexpectedly good at coding. So good, that one vision of the future is that everyone becomes a coder for their own needs, just by talking with their favorite friendly, ubiquitous chatbot. Reed Albergotti explores what that future might look like.

AI makes everyone a coder

The Kaggle report on AI trends

Kaggle is a beloved machine learning competition site and online community. It recently sponsored an essay contest focused on the present and future of AI. Here are the winning entries.

Kaggle essays on AI

Anthropic tackles LLM “explainability”

A persistent criticism of AI is that it functions as a “black box”, delivering results without showing us how it got there. OpenAI competitor Anthropic has released a paper that takes a significant step toward making the functioning of LLMs more understandable. It turns out that LLMs develop abstractions about the data they are trained on, and these abstractions are embedded in groups of interconnected neurons. Anthropic calls these cooperatively-functioning neurons “features’, and believe that they are a key to understanding, and ultimately predicting, how LLMs will produce results.

Decomposing Language Models into Understandable Components

AI detects schizophrenia from speech patterns

Scientists at the UCL Institutue for Neurology have developed new tests, based on large language models, that can objectively detect schizophrenia in patients based solely on their speech patterns. Current methods of diagnosis involve a great deal of subjectivity, and are based on open-ended interviews with both patients and their families.

AI can detect schizophrenia from speech patterns

Google takes aim at AI for Medicine

Google has announced a formidable new suite of AI tools for healthcare, including its AI-powered Vertex search engine, and its Med-Palm-2 multimodal LLM, which was designed and trained specifically for biomedical applications. Google seems to be setting its sights on the biomedical sector as a promising area for product development.

Google takes aim at AI for medicine