Hi all(emaal),
Hope you’re all doing well! Writing to you from the lovely Krakow this time, for those unfamiliar with the city I’d highly advice checking it out, beautiful (walkable!) city with quaint cafes and nice folk.
On to the news-
US congress hosting meeting with most influential tech company CEO's to discuss AI regulation
US considering mandate to force cloud computing firms to report information about their customers to the US government, to 'identify potential AI threats ahead of time'
First public use of Chatbot for help in writing legal summaries for judge verdict
New poll by the AI Policy Institute surveyed 1118 US citizens across various demographic axes on AI perspective, showcasing a general distrust of companies self-regulating in contrast to general US policy
US immigration relying on machine translation more frequently due to work shortages, resulting in an increase in rejections for asylum due to small translating errors
Intelligence agencies like the CIA and FBI jumping on the AI/Large Language model train in order to gain insights from their vast amounts of (official and unofficial, national and international) data
Medium joining an increasingly long list of media outlets that ban the scraping of their websites by AI bots that use it to create new AI models
Guitarist using an AI saxophonist to riff with, with the AI responding to his riffs in kind
Article describing the murky field of generative AI, and artists getting their fair share (if any) of made profits from models trained on their work
Artists call on US congress to ban companies from copyrighting AI-generated art
First dual-degree in AI and medicine, taking 5 years and giving students both a MSc in AI and a doctor of medicine degree, launching at University of Texas
UNESCO called on governments to regulate the use of Generative Artificial Intelligence (GenAI) in education and research, including age limits for users, and published a 'guidance report' on the use of generative AI in education and science
Some universities are starting to drop AI detection software due to concerns of false accusations
Amazon investing up to $4B in Anthropic (known for the Claude model, and competitor to OpenAI)
Microsoft close to rolling out its 'AI Copilot' across its ecosystem (from applications like Word and Excel to one integrated into Windows 11 itself), expected to be made public in November
Google integrating its Bard chatbot into its applications and services (including its cloud storage)
First humanoid robot factory within the USA being built for warehouse robots
Apple's former chief design officer and OpenAI, with backing of SoftBank, in talks to create an 'iPhone for AI', with a 'more intuitive' interface less reliant on screens in order for users to interact with AI
Workday, an enterprise management platform vendor, further putting AI into every aspect of its offerings, up to letting AI generate employee work plans
New Google Deepmind model for the classification of 'missense' mutations, a costly manual operation when done by humans (due to which only 0.1% of mutations have been confirmed to be harmful or not)
Singapore's applied science approach to integrating AI/ML into the public health sector
GPT-4 Vision (GPT-4V) was announced, adding voice and image recognition after a year of safety development
OpenAI revealed DALL-E 3, its new image generation model integrated directly into ChatGPT
New semi-open source voice cloning/generation model released by Coqui called XTTS
Google published the Open X-Embodiment Dataset and their RT-X 'general robotics' model, setting steps towards foundation models in robotics and scaling up results through more data/compute comparatively to LLMs
New superlarge dataset called CulturaX (6.3T tokens, 27TB large)
GPT4-V(ision) paper was published, largely an overview paper on its output and limitations rather than anything about its infrastructure/dataset
Older but still very relevant research article on viewing jobs as collections of tasks wrt exploring and exploiting AI/ML for automation and/or augmentation
Visualising individual neuron's representations in vision transformers
Meta's new text-to-image model Emu made effective through finetuning with a small but high quality dataset of 'extremely visually appealing images'
Research article on how Large Language Models are perfect censorship tools, and there's no real known way of stopping that from happening
Nature survey of 1600 researchers on the effect of AI on/within research
Scientific article commenting on the increasing strain within the scientific community due to a fast increase in quantity of papers whilst amount of practising scientists seems to be stagnating
See you in the next one!
Leven is mooi
