- AI Tangle
- Posts
- ☕️ Microsoft is launching AI-ready PCs
☕️ Microsoft is launching AI-ready PCs
With Microsoft's Build developer conference just a little while away, the tech giant gives the world a peek at the future of AI PCs with its latest line of devices fitted with Qualcomm AI chips. Meanwhile, Scarlett Johansson demands OpenAI to explain why the recent Sky voice from "Her" sounds scarily like her, and the UK's AI Safety Institute picks up speed with worrying LLM jailbreaking results. Join us at AI Tangle as we untangle this week's happenings in AI.
THE BIG AI STORY
Microsoft is launching AI-ready PCs, including a Surface Laptop and Surface Pro tablet, powered by Qualcomm chips capable of running AI tasks offline with a distinctive "Recall" feature. Satya Nadella, CEO of Microsoft, says the Recall feature will be able to search through a log of previous actions on PCs by taking constant screenshots of the user's screen. However, this seems to be a controversial approach to generative AI with concerns about computer storage, so it remains to be seen if Microsoft pushes through with this idea.
Elaborating more on the devices themselves...
As mentioned, the chips powering these devices will be of Qualcomm's make with emphasis on a longer battery life. Copilot+ PCs will also support local execution of some AI models as the tech giant looks to compete with Apple's Arm-based chips and enhance its Windows lineup. Copilot+ PCs will start at $999 and will run Microsoft's Copilot AI chatbot, available for pre-order as of this Monday, with shipping a little later in June. Satya Nadella is confident that "in 2024, AI will become a first-class part of every PC."
5 QUICK HITS
Scarlett Johannson, one of America's most renowned actresses, recently made a statement about OpenAI's Sky that sounds "shockingly" similar to her. She states that OpenAI approached her last September and asked her to voice Sky, with Altman having contacted her agent as recently as two days before the company first demoed Sky's voice, asking for her to reconsider. Despite turning OpenAI down, Johansson felt that OpenAI used her voice anyway, which "forced her to hire legal counsel." Since then, Altman released a statement of his own, stating that OpenAI is pulling the voice of Sky out of respect for Johannson.
Slack has recently been under fire from users upset over the company's confusing terms around using customer data to train its AI services. A viral post made on Hacker News, a social news website around computer science, highlighted that Slack opts users in by default for AI training purposes, requiring them to email to opt out instead. Though Slack claims it doesn't train large language models on customer data for Slack AI, the broader usage of customer data for "global models" immensely lacks clarity on the company's AI efforts.
The AI Safety Institute (AISI), a UK-based AI safety organization established in November 2023, recently published its findings two days before a global AI safety summit in South Korea's capital, Seoul. The findings were about five unnamed but publically in-use LLMs whose guardrails the researchers found were "highly vulnerable." The AISI found it quite simple to make the LLMs talk about denying the Holocaust, being sexist, or encouraging suicide, the guardrailing of which the AI safety summit in Seoul will aim to address.
Building on the previous article, the same UK AI Safety Institute (AISI) has also reportedly opened an office in San Francisco to tackle the rising risks of AI. Officials claim that San Francisco was picked due to its proximity to major players in the field, such as OpenAI, Anthropic, Google, and Meta, all of which are at the forefront of the competition. Aside from the study talked about earlier, one of the AISI's most notable developments was the release of Inspect, a set of tools for testing the safe of foundational models, earlier this month.
BlackRock CEO Larry Fink recently revealed that the asset management giant is in discussions with various governments about financing critical investments required to power the coming AI revolution. Fink warned the G7 nations currently lack sufficient power supply and says that trillions in funding will be needed to meet AI's demand for semiconductor plants and data centers. BlackRock is currently exploring ways to get private investors on board to help tackle this, as governments alone can't afford to fund it.
4 AI TOOLS
Brilliant - AI has become an essential tool - and knowing how it works has become essential, too. Brilliant helps you master AI concepts in minutes a day with thousands of interactive, easy-to-follow mini-lessons. Try Brilliant with a 30-day free trial. *
Mailmodo - Create on-brand email templates in a minute from Mailmodo AI's wide selection of templates to choose from with auto integration for products and coupons.
Shortwave - Organize your inbox the way you want - Shortwave helps you streamline email management with its AI assistant by analyzing your email history to help you write content drafts and pick out what's important.
Persuva - Persuva aims to take your marketing and sales campaigns to the next level, allowing teams to easily craft tailor-made ads that speak to your audience as you know them.
* indicates a sponsored tool
AI READ & WATCH
The Need For UBI (2-min read)
The godfather of artificial intelligence, Professor Geoffrey Hinton, warned people during the BBC Newsnight that a massive benefits reform in the form of universal basic income will be needed to keep up with the inequality that AI will create.
What's Happening At OpenAI? (18-min watch)
Wes Roth, a YouTube channel dedicated to covering AI news, goes over the foggy situation at OpenAI about the Superalignment team, Jan Leike, the company's safety-minded people's dwindling trust in Sam Altman and the "Treacherous Turn."