The Taxman is Using AI Against You

Good morning.

In a past life, I was a fraud investigator, and a part of this investigation is doing OSINT, or “Open Source Intelligence.”. Basically, it entailed looking for information on a “person of interest””.

Yes, i was paid to creep around people’s social media profiles. It’s a thing.

I would always be amazed at what kinds of information I could collect from social media profiles or public company information. And if the person of interest didn’t document their life, it would be a friend or a family member.

Now add some AI to the mix, and I have some concern about where the AI evolution is going. And our lead story is the perfect illustration.

With 5% fewer auditors, the tax department in New York State reported 56% more audits. They’re using AI to target high-earners, specifically those who, during the pandemic, moved from a high-tax state like New York to a low-tax state like Texas. AI is used to comb through these people’s cellphone records to see if they have effectively moved to this state.

It’s not exactly George Orwell’s 1984 yet, but it’s getting there.

The world is generating data “en masse”. We even do it ourselves, taking selfies or pictures of the children and revealing all kinds of data.Our cellphones tell our governments where we are and what we are doing.

Welcome to the Blacklynx Brief

AI NEWS
Good … and Evil

• But while AI can be used for nefarious purposes - at the same time researchers at the University of Cambridge have developed an AI method that significantly expedites the discovery of potential treatments for Parkinson's disease. The AI sifts through millions of compounds to identify those that can prevent harmful protein clumping, refining its findings through lab testing in an iterative process. This approach has already unearthed five compounds far more effective than existing treatments, boosting screening speed tenfold and slashing costs by a thousand times, promising a new era of medical breakthroughs.

• OpenAI has recently fired researchers Leopold Aschenbrenner and Pavel Izmailov, accused of leaking confidential information. Aschenbrenner was linked to OpenAI's chief scientist Ilya Sutskever, and the leaked details reportedly include a rumored AI breakthrough named Q*. This incident hints at ongoing internal issues at OpenAI, even after a period of relative calm following leadership disputes in 2023.

• Apple is set to upgrade its entire Mac lineup with the new M4 chip family, focusing on enhanced AI capabilities. The M4 chips, which include entry-level Donan, high-performance Brava, and top-end Hidra, are expected to start appearing in products from late 2024 to early 2025, beginning with iMacs, MacBook Pros, and Mac Minis. This update comes as Apple aims to boost Mac sales and assert its position in the AI technology race, with more details anticipated at their developer conference on June 10.

• Elon Musk's xAI has launched Grok-1.5 Vision, a significant update to its open-source model, now capable of processing visual data such as documents, photos, and charts. This model, which outperformed competitors like GPT-4V on the new 'RealWorldQA' benchmark, is set to be released to testers and existing users soon. This development indicates xAI's commitment to advancing in the AI field, leveraging Musk's extensive data resources from X and Tesla.

• A new report has surfaced that Adobe's Firefly AI image generator used AI-generated images, including those from rival Midjourney, in its training data. This comprises about 5% of the total data used, despite Adobe's claims of relying primarily on licensed images from its own stock library. The use of competitor's images, which Adobe says underwent moderation, has sparked internal debate over the ethics of such practices. This revelation challenges Adobe's ethical positioning of Firefly and could potentially damage trust with its artist and enterprise customers.

• Researchers from Google DeepMind and the University of Cambridge have successfully taught miniature humanoid robots to play football, demonstrating significant advancements in AI-driven skill learning and adaptability. By training the robots to enhance their capabilities against increasingly proficient versions of themselves, the AI-enabled robots exhibited substantially faster movement and recovery from falls, along with the ability to autonomously develop strategic behaviors. In this stage they still look cute.

• Slightly less cute : Boston Dynamics has introduced a new, all-electric version of its Atlas humanoid robot, which is sleeker and designed for practical use. This robot, now without exposed cables and featuring a ring-light head, demonstrates enhanced agility and dexterity, such as rising from the ground and rotating its torso fully. In collaboration with Hyundai, this robot will undergo real-world testing in automotive manufacturing, showcasing Boston Dynamics' continued innovation in robotic design beyond traditional human capabilities.

  

• Adobe has announced a significant update to Adobe Premiere Pro, introducing advanced AI tools that revolutionize video production. These new features include extending video clips, editing objects within scenes, and generating B-roll footage via simple prompts. Additionally, Adobe is developing a new AI model, Firefly Video, and plans to integrate third-party AI models from companies like OpenAI and Runway. The updates, which also include AI-enhanced audio editing capabilities set for release in May, position Adobe at the forefront of AI-powered video editing, ensuring it remains a preferred platform by offering a wide range of tools and options.

• Microsoft has invested $1.5 billion in G42, an AI company based in Abu Dhabi, gaining a minority stake and a board position. This strategic move allows G42 to use Microsoft's cloud technologies for AI development, amidst a backdrop of U.S. concerns regarding G42's previous ties with China. To alleviate these concerns, G42 has agreed to exclude Huawei from its systems and comply with strict security protocols. This partnership not only mitigates geopolitical tensions but also positions Microsoft favorably in the Gulf region's expanding AI market.

• AMD has launched its latest Ryzen PRO mobile and desktop processors, positioning itself as a strong competitor in the AI-enhanced processor market. The new Ryzen PRO 8000 desktop processors are being marketed as the first AI-enabled processors for business use. On the same day, Nvidia introduced two new AI-focused GPUs, the RTX A400 and RTX A1000, designed to enhance AI capabilities across various computer systems. This simultaneous release underscores the ongoing competition between AMD, Nvidia, and Intel in the tech industry, further highlighted by the familial link between AMD’s CEO Lisa Su and Nvidia’s CEO Jensen Huang. Interesting conversations at the dinner table for sure.

Closing Thoughts

That’s it for us this week.

Very interesting week with significant developments in especially robotics.

Have a nice weekend.

If you want a version of this newsletter with pictures etc. Be sure to subscribe below:

https://www.blacklynxbrief.com/subscribe