AI Highlights

My top-3 picks of AI news this week.

Tesla
1. Musk’s robo-revolution

Tesla made a serious impact at their “We, Robot” event this week by introducing new tech aimed at revolutionising transportation and robotics. Here’s the signal:

• Robovan: An autonomous electric vehicle designed to transport up to 20 people and goods in high-density areas.

• Cybercab: An autonomous taxi expected to be available by 2026-2027, priced under $30,000.

• Tesla Bot: Humanoid robots priced between $20-30K, intended for performing household tasks with full self-driving capabilities. It can even talk fluently to humans.

Alex’s take: Musk has always been ambitious with his timelines—whilst they’ve historically faced delays, having a Cybercab pick me up in 2026 does sound exciting. Competition is intensifying in the autonomous driving space, with established players like Waymo and Cruise already having quite the foothold. I’m curious to see how this plays out.

OpenAI
2. OpenAI launches Swarm for multi-agent orchestration

Swarm is an experimental framework by OpenAI that allows multiple AI agents to work together. It makes it easy to control and test how these agents interact and coordinate their tasks, which can be helpful for handling complex situations that require different skills or tools.

Key points:

• Agents: These are like individual bots with specific instructions and tools to complete tasks. If needed, an agent can pass its task to another agent (called a “handoff”).

• Handoffs: This is when one agent passes a task or conversation to another agent that might be better suited for it.

Alex’s take: While Swarm is mostly for educational purposes, I think it’s a nice opening for developers who want to experiment and understand how multi-agent systems work. Once again, the barriers are lowering for anyone to create scalable AI products—now with agents.

Nobel Prize
3. Geoffrey Hinton awarded Nobel Prize in Physics

Geoffrey Hinton, often hailed as the “godfather of AI,” has been honoured with the Nobel Prize for his work in artificial neural networks. His innovations have laid the groundwork for many of the AI technologies that power today's applications, from virtual assistants to autonomous vehicles.

Alex’s take: Let’s just say the reception has been mixed, not for the foundational work of Hinton but for its categorisation. Many have argued that this is not physics—the science that seeks to understand the principles and dynamics of the physical universe—but more suited instead to computer science or a Turing prize.

Something else that interests me is his public endorsement praising Ilya Sutskever for his role in the decision to fire Sam Altman from OpenAI. What does Hinton know that we don’t?

Content I Enjoyed

Machines of Loving Grace

Dario Amodei is the CEO of Anthropic. He recently wrote a ~15,000-word essay highlighting key areas where AI could dramatically improve human life. As a techno-optimist myself, I thought I’d share my 3 key takeaways to spare you 30 minutes.

• AI could compress 50-100 years of medical progress into 5-10 years. Amodei also sees potential for eliminating most infectious diseases, cancer, genetic disorders, and Alzheimer's, as well as doubling the human lifespan to ~150 years.

• AI could accelerate economic growth in developing countries and potentially raise sub-Saharan Africa’s GDP to levels comparable with China within a decade. There is potential for 20% annual GDP growth in developing nations.

• AI will have a vast impact on our economy. In the short-term, comparative advantage will keep humans relevant and productive. In the long-term, AI may become so effective and cheap that our current economic model becomes obsolete.

With the mainstream media painting the doom-and-gloom picture almost daily, it’s refreshing to think beyond the current challenges and look forward to a future where AI brings profound positive change to humanity.

Idea I Learned

Are we prepared for the future of data privacy?

Two Harvard students, AnhPhu Nguyen and Caine Ardayfio, have developed I-XRAY.

It exposes personal data in public spaces by connecting Meta’s Ray-Ban smart glasses to a facial recognition system.

The user looks at a stranger, and the glasses detect when someone is in the frame.

Then, a video stream is sent from the glasses to Instagram Live, where a computer then carries out a reverse face search of the person in the frame.

LLMs come into play to find the person’s name from their likeness. But it goes a step further.

It can find someone’s phone number, home address, and family members.

Something I must add is these students made I-XRAY not for misuse but as a thought-provoking experiment.

Whilst they’re not releasing the technology, it does make you wonder what large corporations who have a near-unlimited money pot are doing right now.

Privacy is more important than ever before in the age of AI—it’s our job to stay on top of what were share and manage our digital footprint.

The students shared a great supporting document which outlines how you can remove your information from data sources like FastPeopleSearch and Pimeyes—which essentially renders technology like I-XRAY ineffective on you.

You can access it here.

Quote to Share

AI Hawk's impact on job applications in 2024:

“I applied to 2,843 Roles.”

1. Thousands of people are using AI to automatically apply for jobs at scale

2. Some users report applying to over 2,800 jobs in 3 months using AI tools

3. This creates a feedback loop where AI is writing cover letters for other AI to read

This trend emerged in late 2023/early 2024. AI Hawk has gone viral on GitHub, with thousands of users and a Telegram community of 4,700+ members. Some report getting interviews within days of mass-applying via AI. It's wild to see how quickly this has transformed the job application landscape.

Source: Article from 404 Media

Question to Ponder

Trust is everything in healthcare. And AI still has a way to go to have the necessary reliability for patient care. However, AI can be a powerful tool for good if leveraged correctly.

Firstly, the human-AI collaboration. AI should be used as an augmentation tool rather than a replacement. The final decision should always lie with the expert human in the loop.

Secondly, validating AI recommendations. AI can be particularly useful in helping with initial diagnostics or administrative efficiency, but qualified healthcare professionals must validate AI-generated recommendations to prevent errors or hallucinations like you mentioned.

Thirdly, transparent data provenance. We’ve got to leverage trusted, peer-reviewed and well-documented datasets. Any AI model used in healthcare should be trained only on the highest quality dataset to build confidence in its output.

The upside to this is promising. We’ve already seen AI for medical reasoning outperform doctors. Google’s AMIE, released in January 2024, outperformed both doctors alone and doctors equipped with AMIE for diagnostic accuracy. What’s more, Google’s Med-Gemini results have been preferred over those of human experts on long-form text-based tasks, including medical summarisation, referral generation, and medical simplifications.

Right now, AI can reduce workloads, reduce oversights, and assist in complex decision-making. We have a way to go before we can trust and rely on its decisions outright. It’s all about combining the power and scale of AI with human expertise to reach better patient outcomes.

In the future, I foresee a world where AI makes most of the decisions—from diagnosis to patient care.

💡 If you enjoyed this issue, share it with a friend.