There was the usual flurry of product announcements this week from various AI companies (e.g., OpenAI GPT-4o model update, Mistral’s product updates, DeepSeek’s reasoning model, Perplexity’s launch of its new online shopping feature).

But I thought it would be more intriguing for you, my readers, to experience a broader, yet curated mix of what’s happening in the AI landscape. So let’s dive right in!

#1 NEO (But Not That Matrix Guy)

Autonomous agent sets new benchmark in automated ML development

Meet NEO, a new AI agent that after two years of quiet development has emerged as something quite special — an AI that can actually build and deploy other AI systems.

For those not in the data science world, earning Kaggle Grandmaster status is the equivalent of winning MasterChef while solving a Rubik's cube blindfolded. NEO has done just that, winning medals in 26% of Kaggle’s (a machine learning and data science community) competition entries, handily beating the previous record of 16.9% held by OpenAI’s state of the art o1-preview model.

In various tests, users showed off NEO’s capabilities by building fraud detection systems and analysing book reviews — tasks that can take teams weeks or months to complete. What's clever about NEO is how it works — imagine a group of specialised AI agents working together like a well-oiled machine, each handling different parts of a project and learning from experience. (Curious about how such AI agents work? Check out my recent piece, AI Gets Promoted.)

Perspectives:

While NEO's Kaggle success certainly turns heads, the real test lies ahead. Winning data science competitions is one thing; handling messy real-world data with complex business constraints and legacy systems is quite another.

However, NEO's emergence is timely, coming into a market where AI talent is scarce and companies are struggling to build their machine learning (ML) capabilities, NEO's ability to automate complex ML workflows could democratise access to sophisticated AI development. This could help to bridge the existing data science talent gap and also reshape the role of ML engineers — shifting their focus from routine implementation to higher level innovation and overseeing ML systems.

#2 AI's Double Edge

Study reveals AI's impact on freelance jobs and some surprising results

A groundbreaking study analysing 1.3 million online freelance job posts has revealed just how quickly Generative AI is reshaping the employment landscape. The findings paint a clear picture: while traditional automation took years to impact industries, but AI's influence is unfolding in months.

After ChatGPT's debut, job posts for writing-related work dropped by 30%, while posts for software development and engineering positions declined by 21% and 10% respectively. In the visual arts, the arrival of tools like DALL-E and Midjourney led to a 17% decrease in posts for graphic design and 3D modeling freelancers. And unlike previous technological disruptions, there's no sign of these trends reversing.

But it's not all doom and gloom. The remaining job posts are becoming more sophisticated, with employers requiring 2.18% more skills per posting. Not only are they asking for more complex skill combinations, but they're also willing to pay 5.71% more for this expertise — suggesting a shift toward higher-value work that combines human judgment with AI capabilities.

Perspectives:

The data paints a nuanced picture of an emerging "AI-human hybrid" job market. While raw numbers show declining post volumes (for now), the increased complexity and higher pay for remaining posts suggest a shift toward roles that blend human expertise with AI proficiency. Still, its early days and only time will tell how AI will fully impact jobs and the future of work.

#3 Lights, Camera, Algorithm!

Hollywood's newest studio strives to balance tech with talent

Promise, the latest kid on Hollywood’s blocks, launched with significant backing from Peter Chernin's North Road Company and Andreessen Horowitz, the VC fund.

Led by former Fullscreen CEO George Strompolos, ex-YouTube executive Jamie Byrne, and AI artist Dave Clark, the studio aims to integrate AI throughout the film and series production process. The studio is developing proprietary software called MUSE that will incorporate Generative AI tools across the entire production pipeline.

The founding team positions Promise as "a new studio built from the ground up to push the boundaries of storytelling with the transformative power of Generative AI." They draw an parallel to YouTube's impact, noting how it democratised content distribution, while suggesting that Generative AI will democratise content creation.

In their launch letter however, they're careful to emphasise that humans remain central to their vision — the technology may be the backbone, but the creative community, as they put it, is still "the heart and soul."

And speaking of that creative community, Promise is already reaching out to both traditional Hollywood talent and AI artists to develop their initial slate of projects.

Perspectives:

The timing of Promise’s launch is noteworthy, landing just as Hollywood emerges from a year of intense negotiations about AI's role in entertainment. With industry bodies, the Writers Guild of America and SAG-AFTRA, having recently secured new agreements about how AI can be used in productions, alongside the emergence of high fidelity AI video generation models such as Runway’s Gen-3 Alpha, OpenAI’s Sora and many others, Promise will likely need to walk a tightrope to avoid alienating a creative community that is already on edge.

Promise's comparison between YouTube's impact on content distribution and AI's potential effect on content creation also warrants closer examination. While YouTube primarily addressed technical hurdles — essentially providing a platform where anyone could upload their videos — Promise faces a far more intricate challenge. High quality film creation involves a complex multi-stakeholder creative processes, technical expertise, and substantial capital investment. AI tools alone will not democratise the ability to craft compelling narratives or produce such content.

#4 Ignite 2024: Microsoft Brings the AI Heat

Copilot Actions leads wave of enterprise AI updates

Much like Google I/O and Meta Connect earlier this year, Microsoft's Ignite 2024, its annual enterprise technology conference, has AI written all over it.

While the company made several significant AI announcements, Copilot Actions emerged as the headliner. This new feature automates repetitive office tasks, teaching Copilot to handle routine workflows — from summarising meetings to generating reports — using simple fill-in-the-blank prompts.

Microsoft is also integrating these capabilities into SharePoint, allowing organisations to create custom AI agents for managing and querying their document repositories. If you've read my recent piece, “AI Gets Promoted”, you'll recognise this as part of the broader industry shift toward AI agents that are capable of not just acting as assistants, but independently carrying out assigned tasks.

Office 365 continues to receive meaningful AI updates. Excel will get intelligent templates to jumpstart spreadsheet creation, PowerPoint will translate presentations into 40 languages, and Outlook is being enhanced with AI-powered scheduling.

Teams is getting a babel fish-worthy upgrade (IYKYK!). The new interpreter feature will let you chat in your preferred language while your colleagues hear it in theirs, complete with your voice. For multinational organisations managing teams across language barriers, this could transform how global colleagues collaborate, though we'll need to wait until early 2025 to see it in action.

Meanwhile, on the security front, Microsoft is launching Zero Day Quest, offering $4 million in bounties for those who find AI and cloud security flaws.

Perspectives:

Microsoft's $4 million bug bounty program for Zero Day Quest makes a lot of sense and is well-timed. AI threats of all sorts have been appearing on the horizon and enterprise IT teams are on high alert.

Yet even as Microsoft does a decent job selling the security story, they're swimming in complexity soup. Sure, Copilot Actions and the Wave 2 updates from a couple of months back are solid, but Microsoft has a knack for turning simple ideas into complex feature lists that make even tech leads scratch their heads. Until they crack the code on making these tools intuitive and robust, they're at risk of creating the world's most sophisticated shelf-ware — an all-too-familiar story in enterprise tech.

#5 AlphaFold's Half-Open Door

World's best protein modelling AI now available, with strings attached

AlphaFold 3, released in May of this year, is the latest iteration of DeepMind’s revolutionary protein folding model. Going beyond its predecessor’s ability to predict protein structures, AlphaFold 3 can simulate intricate interactions between proteins, DNA, RNA, and small molecules, fundamental processes that make life possible.

This broader capability opens new paths for understanding cellular processes, from gene regulation to drug metabolism, at a scale previously out of reach. And Google DeepMind has just released AlphaFold 3's source code to the academic world.

DeepMind's approach to balancing academic access with commercial interests takes a careful middle path. The code is freely available under a Creative Commons license, but those crucial model weights? They require explicit permission for academic use. Commercial applications, particularly in drug discovery, remain firmly off-limits.

In the meantime, competing model versions — based off specifications described in DeepMind’s original paper — are making it to market, including from Chinese tech giants Baidu and ByteDance (yes, the TikTok folks), and smaller players such as San Francisco-based Chai Discovery and Ligo Biosciences.

Perspectives:

DeepMind's ‘partial open source’ approach with AlphaFold 3 reflects a growing tension in AI development between scientific transparency and commercial advantage. The emergence of alternative implementations — while not yet as good as Alphafold 3 — suggests this restrictive approach may ultimately be self-defeating.

While commercial restrictions currently limit pharmaceutical applications, the academic research enabled by this release will continue to advance understanding of disease mechanisms. Moreover, it seems only a matter of time before these models become fully open-sourced or commercially licensed, given the competitive pressure from multiple implementations and the broader industry trend toward openness.

Justin Tan is passionate about supporting organisations and teams to navigate disruptive change and towards sustainable and robust growth. He founded Evolutio Consulting in 2021 to help senior leaders to upskill and accelerate adoption of AI within their organisation through AI literacy and proficiency training, and also works with his clients to design and build bespoke AI solutions that drive growth and productivity for their businesses. If you're pondering how to harness these technologies in your business, or simply fancy a chat about the latest developments in AI, why not reach out?