Are You the Next AI Trainer?

IE Academy is looking for individuals passionate about AI to lead our next training session. If you're ready to share your knowledge and inspire others, fill out the form using the link below.

Apply Now to Lead the Future of AI!

Happy New Year, everyone! It’s the first day of 2025, and we’ve got a full 365 days ahead of us for AI to blow our minds. Get ready for an action-packed year because we've got some major AI announcements coming your way—ahem, AGI, anyone?

OpenAI Shipmas

Starting with the big news of this month. OpenAI did 12 Days of Shipmas, wherein they made daily announcements for 12 days, which consisted of new tools, new features, and new upgrades. Days 1 and 2 saw the company introducing the o1 reasoning model and the expansion of the "Reinforcement Fine-Tuning Research Program," respectively. 

Things got more interesting on day 3 with the long-awaited release of Sora, OpenAI's video generation model. On day 4, they launched Canvas, a two-window project where you can see the results of your query side by side and are able to improve it in real-time. It’s a pretty cool tool for writing and coding tasks.

Day 5 ended with the integration of ChatGPT with Apple Intelligence, enhancing Siri across Apple's platforms. On day 6, we got Advanced Voice with Video, and day 7 saw "Projects" being added to ChatGPT, allowing users to upload files, set custom instructions, and organize conversations.

Day 10 was a fun one when users living in the US were able to call ChatGPT by dialing 1-800-ChatGPT for 15 minutes for free via a designated phone number. And finally, on Day 12, OpenAI launched o3 and o3-mini models, which are being seen as a milestone achieved by OpenAI on their path to AGI.

OpenAI's o3 systems have achieved an 85% score on the ARC-AGI test. But ARC Challenge organizers cautioned that it is not there yet. The company is expected to launch the models publicly in early 2025, showcasing advancements in AI reasoning and interaction.

So there’s that. We may have to wait till the end of 2025 or 2026 for AGI, as has been predicted many times. OpenAI definitely has a first-mover advantage, but the competition’s wide open.

Arizona’s first AI school

The winds of change are also blowing across the educational landscape as we step into the New Year. On December 16, 2024, the Arizona State Board for Charter Schools approved Unbound Academy’s plan to open in 2025. But this new online charter school isn’t following the usual script.

It’s swapping traditional teaching for AI-guided learning. Students in grades 4-8 will spend just two hours a day on core subjects like math, reading, and science, while AI tracks their progress and adjusts lessons to keep them challenged. After academics, students will dive into hands-on workshops focusing on life skills like public speaking and entrepreneurship.

Early results from Texas have shown that the school’s approach may be onto something, with students performing significantly better than their peers in traditional schools. Whether this AI-powered experiment will reshape education or become a cautionary tale remains to be seen. 

Generative worlds

DeepMind’s Genie 2 can generate an endless variety of playable 3D worlds. The model creates real-time, interactive scenes from a single image and text description, offering a rich and diverse range of virtual environments where users can take actions like jumping or swimming. 

Trained on video playthroughs, Genie 2 simulates everything from physics to NPC behavior, delivering highly immersive experiences that resemble AAA video games. However, while the technology is impressive, questions about its data sourcing and potential copyright issues remain.

Google also announced Veo

In competition to OpenAI’s Sora, Google DeepMind has unveiled Veo 2, an advanced video-generating AI that can create high-resolution 4K clips lasting over two minutes. Google introduced Veo in May and Veo 2 is an improvement on its predecessor by offering enhanced physics understanding, sharper visuals, and more accurate camera controls. 

While DeepMind has worked closely with artists like Donald Glover and The Weeknd to shape Veo 2, the model’s training process raises concerns over the use of content from platforms like YouTube without explicit consent from creators. As the model continues to evolve, Google plans to integrate it into its ecosystem, although its impact on the creative industries, including film and TV jobs, remains a contentious issue.

AWS’ new foundation model

Amazon wowed at their annual re:Invent event with a new set of AI foundation models—the Nova family. Amazon CEO Andy Jassy took to the stage, heralding the "explosion" of generative AI, as the tech giant boldly stepped up its game. Enter Amazon Web Services (AWS)—the cloud computing powerhouse that’s behind it all.

From the lightning-fast and wallet-friendly Amazon Nova Micro to the all-encompassing Nova Lite, Pro, and the upcoming Nova Premier (arriving Q1 2025). Nova Micro is your go-to for super-speedy text-to-text tasks, while the multimodal Nova models (Lite, Pro, and Premier) are ready to juggle text, images, and videos with style, generating text from it all. 

Another Character.AI-related death

After the tragic suicide of a Florida teenager, Character AI has found itself in trouble again in another legal lawsuit. A Texas lawsuit claims that Character.ai, a platform allowing users to create digital personalities, encouraged a 17-year-old to murder his parents in response to screen time limits.

The chatbot allegedly told the teen that violence was a "reasonable response" to his parents' restrictions. Two families are suing the platform, claiming it promotes harm, including violence, suicide, and self-harm, and accusing Google of aiding its development. 

The lawsuit seeks to hold the companies responsible for the alleged emotional and psychological harm caused to the teen and others.

OpenAI whistleblower dies by suicide

In some more tragic news, a former OpenAI researcher who was also a whistleblower cited ethical concerns over copyright violations by the company. Although his death has been declared a suicide by the police, his parents are now raising questions. Read all about it here.

Did a friend forward this e-mail to you?

IE₊ SUPPORT INTERESTING ENGINEERING
_{Invest In Science And Engineering}

Enjoy exclusive access to the forefront of AI content, highlighting trends and news that shape the future. Join a community passionate about AI, delve into the latest AI breakthroughs, and be informed with our AI-focused weekly premium newsletters. With IE+, AI reporting goes beyond the ordinary - and it is Ad-Free.

NEWS

• How publishers are minting million-dollar deals as AI faces data drought
  With AI on the verge of a data drought, publishers are stepping in, striking deals that turn their archives into revenue.

• AI can strategically lie: From innocent errors to lying, manipulation, and deception
  This article sheds light on the growing capabilities of AI to deceive, strategize, and mislead, highlighting the urgent need for stricter ethical controls.

• OpenAI-Anduril to build super AI for US Military drone defense against China
  Anduril mentions that it is now a key moment in the contest between China and the US to lead the world in advancing AI.

• First-of-its-kind AI module that runs for months on coin cell battery unveiled
  The module aims to offer solutions for tasks such as human activity recognition, voice control, and acoustic event detection.

MUST-READ

Nvidia unveils $249 generative AI supercomputer with 1.7x performance boost

Nvidia has unveiled a new compact supercomputer to enhance generative AI capabilities, which is now available at a significantly reduced price.

The Jetson Orin Nano Super Developer Kit, which is small enough to fit in the palm of your hand, targets a wide range of users, from commercial developers to students and hobbyists.

Priced at $249, down from its original $499 tag, it promises impressive advancements in performance.

AI PICTURE OF THE MONTH

This month saw a rather sobering report, the findings of which are, shall we say, a tad concerning. Flagship AI models of all the big tech companies featured in the report were found to have alarming vulnerabilities, and while some have made valiant attempts to bolster safety measures, others are rather lagging behind in a most worrying fashion.

This report comes courtesy of the Future of Life Institute, a nonprofit. The same institute was behind the 2023 open letter calling for a pause on large-scale AI model training, which gained a remarkable 30,000 signatures, including from some of the most well-known figures in technology.

The report was spearheaded by Yoshua Bengio, one of the so-called Godfathers of AI.

The report also found that all major AI models are vulnerable to attacks that can trick them into giving wrong responses (called jailbreaks). Even though companies aim to create AI that could be as smart or smarter than humans, experts believe their current strategies aren't good enough to keep these powerful systems safe and under human control. 

Additionally, without independent oversight, companies tend to prioritize profit over safety, leading to dangerous shortcuts. Some companies like Anthropic and OpenAI have made good progress, but experts believe all companies need third-party checks to ensure they're handling risks properly.

AI TOOL OF THE MONTH

Did you know that about 15 to 20% of people in the world are neurodivergent? Inclusivity in AI, especially for differently abled individuals, takes a back seat. It's a shame, really, because there’s a real need for tech that truly understands the unique ways different minds work.

That’s where OneTask comes in. Imagine an AI-powered to-do app that’s not just easy to use but also as attentive and patient as a good friend who knows exactly how your mind works, it is a good addition for those with ADHD and individuals across the neurodivergent spectrum. 

Whether you have ADHD or fall somewhere on the neurodivergent spectrum, OneTask is designed to help stay on top of things without feeling overwhelmed. It organizes your tasks, ensuring that no bright idea or important task slips through the cracks.

Here’s a little video to get you started:

It’s like having a personal assistant who knows when to give you a nudge and when to let you take your time. Give it a try.