Emerging AI’s Year End: Roundup for November and December 2024

2024 is behind us and it's time for our bi-monthly AI roundup. It’s also a good time to look back over a year that some believed would see substantial consolidation and cooling in the field, as the public tired of the hype and companies looked for practical applications with direct use cases, enabling prudent decisions on ROI. 

But while there were moments of pullback, 2024 saw the major tech companies continuing to pour in billions, as they both launched AI-powered innovations and spent freely in an infrastructure arms race aiming to build, secure, and power a range of new data centers seen as essential for AI’s growing future. 

One example of an AI-driven windfall is the Texas Pacific Land Corp (TPL), which gained 230% in value by late November, giving a company of 100 a value over $40 billion (larger than Halliburton and four times the size of nearby American Airlines). Offering land and water solutions in a shale-rich region of West Texas, AI has driven its surge—specifically their new data centers.

And some investors, like BlackRock, believe the trend will continue, stating in their 2025 outlook report (per Reuters):

"We stay risk-on ... and go further overweight U.S. stocks as the AI theme broadens out."

Speaking of broadening out, November saw yet another new application of AI in a so-called AI Jesus, deployed by St. Peter's Church's Confessional in Lucerne, Switzerland. 

Trained on the New Testament, it is built to hear confessions and give spiritual advice (and able to do so in ten languages). While it frequently answers questions with more questions, two-thirds of confessional users reported having a spiritual experience, showing success in yet another novel AI application.

And while our round-up doesn’t necessarily aim to give advice, it does aspire to cover the major AI topics from the period. 

If you need to catch up on our prior roundups of AI advancements for 2024 to date, take a look at the links below:

• January and February

• March and April

• May and June

• July and August

• September and October

Working Integrations 

Autonomous AI Agents may be the goal for 2025, but there are already numerous in-progress AI integrations in 2024 to consider.

This month, Wired reported on Botto, an artist who’s made $4 million in sales, who also happens to be an AI. First created in 2021 by German artist Mario Klingemann, entrepreneur Simon Hudson, and computer scientist Ziv Epstein, Botto makes use of a taste filter curated by a voting community of enthusiasts. And it’s now getting an upgrade to allow chatting and access to a knowledge base, in the form of a modified version of a Mistral open source LLM. 

The Wall Street Journal reported on AI integrations in the food industry at famous consumer brands, from Taco Bell to Oreo cookies. 

Yum Brands (owner of Taco Bell, Pizza Hut, and KFC) has been using AI across the board, but reports success, especially from AI-powered marketing pilots they claim are increasing sales and reducing churn. 

Specifically, they’ve been taking advantage of AI for personalization, delivering emails customized to the individual, taking account of time of day and purpose (upsell, retention, referral, win-back). 

The system can tell the difference between users who order for special events versus those who like to try new things across the menu. It selects from pre-written templates and decides when and how to send. 

They aim to grow from this to offer customized messages at kiosks and drive-throughs, too, that are unique to location. 

Meanwhile, at Mondelez International, machine learning is being used to speed up the creation of new recipes, developing them for certain taste profiles. Their integrations mean less time in labs with a faster path to market, reaching trials four-to-five times faster. 

The trick is limiting the recipes using desired traits, such as flavor (“in-mouth saltiness” or “vanilla intensity”), scent, and appearance, while also considering cost, environmental effect, and nutrition. 

At Colgate, digital twins are used to simulate actual customers, with a similar goal of accelerating the development of new and enhanced products. They’re also working with an outside firm on a generative AI solution that analyzes internal customer research to identify trends and allow workers to more easily pull needed information. 

The last two months also saw progress in the integration of AI in financial firms like Citibank, with solutions to help navigate policies and documentation as well as bolster areas like HR, compliance, and risk evaluation. 

Meanwhile, the Salesforce AI integration, Agentforce, has reached the 2.0 stage, with a boom in investor confidence based on demos of new capacities, including improved reasoning and agent integration with Slack. 

Devices and the AI Chip Industry at the Close of 2024 

Despite Intel’s struggles, AI hardware overall has shown continued success over this period and the year as a whole, with none surging more than Nvidia and TSMC, the South Korean firm that manufactures most of their chips. 

Their (still forthcoming) Blackwell chip is expected to be the state of the art, and despite production delays and reports of extreme complexity slowing down suppliers, investment firms like Morgan Stanley remain bullish on the chip for 2025, keeping Nvidia at the top of the charts.  

California chip-maker Broadcom is having its own moment, as the big tech powers seek to diversify from the market-dominating Nvidia. While this includes cloud hyperscalers, it has also featured a Broadcom Apple deal for developing a new AI chip (code-named Baltra and ready for production in 2026). Broadcom also manufactures networking technology for data centers and previously worked with Apple on 5G components for the iPhone. 

With a stock surge and its value reaching $1.2 trillion in mid-December, there’s a sense that the company is witnessing its own Nvidia-like spike based on the expected demand for AI components needed in data centers. 

Amazon is also nearing roll-out of their own custom AI chips, including the Trainium 2, which is already being tested by Amazon-bolstered Anthropic. While the move is aimed at reducing costs and diversifying the field, Amazon’s cloud division continues to make heavy use of Nvidia chips.

As AWS’s vice-president of compute and networking services Dave Brown told Ars Technica:  

“We want to be absolutely the best place to run Nvidia… But at the same time we think it’s healthy to have an alternative.”

We reported in a prior roundup on the nuclear push to power these energy-hungry data centers, and Meta joined the ranks of Microsoft, Google, and Amazon in this period.

And in terms of interfaces, one of my nephews greeted me over the holidays wearing smart glasses with AI that worked in conjunction with a smartphone. While the glasses didn’t succeed at helping him make moves in a board game, they did take and send pictures of things he was looking at, described settings correctly, and provided a lot of entertainment.

And we should expect a lot more of this in 2025, as other companies join Meta in bringing multi-modal AI to our faces. 

In a partnership with Samsung, Google is leveraging their Gemini AI models both for AR/VR headsets and smart glasses, despite a prior failure with Google Glass. 

The Android-XR driven glasses have been demo-ed getting information on the world around you and chatting with friends, as well as combining with a smartphone to recall things you’ve encountered at certain locations after the fact. 

[For the story on the Harvard students quickly turning Meta’s smart glasses into a facial recognition system, check out this edition of The PTP Report on data safety with AI.]

Innovations Smaller… 

Speaking of Harvard, the prestigious Ivy League school announced in December they were releasing a high-quality data pool including over a million public domain books for use in AI training. More than five times the size of Llama’s comparable data source, this Harvard AI dataset comes highly vetted and curated and is released by their new Data Initiative in partnership with OpenAI and Microsoft. The stated goal is helping startups and leveling the playing field with quality, copyright-cleared content. 

Other intriguing innovations from this period NOT from big tech include:

• AI startup Cleanlab has released a Trustworthy Language Model that grades LLM answers on their reliability. Returning a score between 0 (wholly unreliable) and 1 (completely reliable) for each query, it helps companies flag areas risky for hallucination. It can also get multiple models to try their answers against each other.  

• An AI startup made news in November by releasing a game (free to play online here) that’s entirely generated, in real-time, by AI. Using a system trained on hours and hours of footage of people playing Minecraft, it anticipates what should come next, offers a number of map options, and can even be played using your own custom images in a Beta mode.

• Cognition AI’s software developer application, Devin, was released to the public in December, promising the capacity to automate bug fixes, in addition to other capabilities. 

… and Innovations Bigger 

There were also loads of major, end-of-year releases. But first, here’s an AI year-end review for the big players:

Several new releases hit the public in just the last two months, including reasoning models that use reinforcement learning, like OpenAI’s o1 (full/pro) and Google Gemini 2 AI, as well as Llama 3.3, a mini model that reaches near GPT-4 level capacity at just 70B parameters. 

One look at the Chatbot Arena LLM Leaderboard shows these, and other new models, making waves. 

But with names like Gemini-2.0-Flash-Thinking-Exp-1219, ChatGPT-4o-latest (2024-11-20), Claude 3.5 Sonnet (20241022), and Deepseek-v2.5-1210, it also echoes Wharton School professor (and AI researcher) Ethan Mollick’s lament on model naming, as he posted a request for the companies to help consumers with more descriptive names, and “Please stop naming AI like files on my hard drive!”

Other innovations from the biggest players include:

• AI video is out to the public and off to the races. This includes OpenAI’s long-promised Sora and Google’s Veo 2 leading, with more on the way. Unsurprisingly for anyone who’s tried these, AI technology is increasingly being folded into professional studio productions.  

• OpenAI innovations in 2024 didn’t stop with Sora and o1, as they also dropped news of an o3 model, due early next year. While they skipped o2 for naming overlaps, this new version is supposedly 20% better than o1 at reasoning tasks and, according to CEO Sam Altman, could beat most of their top developers at software tests. This accompanies a stunning $200 price tag per month to harness o1 AI in business applications.

• Clashes are continuing at the big firms, with Elon Musk and Meta both trying to bar OpenAI’s move to for-profit and Google also asking the FTC to crackdown on Microsoft’s OpenAI deal. 

• And speaking of clashes, Wired created an AI copyright case tracker that you can check out here. This shows an updated status on who’s suing who, in all the various AI-driven cases. 

• Google is already using AI to scan your images for data, and you can check out the success (or failure) of the insights it’s gleaning from your pics using a free utility that’s online here. It uses the Google Vision API to demonstrate this process.

• Meanwhile the AI agents are coming, with 2025 expected to see a flood of them from major players. OpenAI’s is reportedly codenamed “Operator” and expected to be in preview in January, according to Bloomberg. 

• Amazon’s introducing its own multimodal AI model family called Nova and also building what may be the world’s biggest AI supercomputer with Anthropic, expected to be available in late 2025. 

• Not to be left out, Apple has widely rolled out Apple Intelligence and is reportedly developing a wall-mounted AI tablet for entertainment, controlling appliances, video calls, and more. 

PTP and AI 

As a major tech recruiting firm, PTP has been involved with AI for years now, helping companies find talent in AI, whether it’s cybersecurity analysts with AI skill, AI ethics specialists, or AI and ML engineers. 

But as we come to the close of 2024, we’ve also got experience working on our own AI integrations in-house, as well as directly helping clients with theirs, making this a busy and exciting time as we gear up for the emerging AI in 2025. 

In Conclusion 

As usual in these roundups, there’s too much AI news to hit, even bi-monthly (with areas like science and medical research, defense, regulation, and international competition all capable of having their own reports). 

In response to all the early speculation about how AI innovations in 2024 would impact jobs, the Harvard Business Review ran a piece in November highlighting the early returns. As you might expect, positions prone to automation saw a 21% decrease in listings (including writers, software developers, graphic design freelancers, and engineers overall). 

But interestingly, those that remained are in higher demand and also feature higher complexity, requiring more skills than their predecessors. This underscores the benefits in upskilling now with AI, regardless of your area of expertise. And this is true both for companies and the workforce alike.

Expect our first roundup for 2025 at the end of February, and until then, keep exploring!

References  

AI Boom Propels a Wild-West-Era Texas Landowner to 230% Stock Rally, Bloomberg

BlackRock bets on AI-driven stocks rally but US debt clouds 2025 outlook, Reuters

Switzerland's AI Jesus answers questions of faith, DW

Botto, the Millionaire AI Artist, Is Getting a Personality, Wired

Taco Bell and KFC’s Owner Says AI-Driven Marketing Is Boosting Purchases, The Wall Street Journal

Oreo Owner Mondelez Taps AI to Tweak Its Classic Snacks, The Wall Street Journal

Toothpaste maker Colgate testing new product ideas on 'digital twins', Reuters

Salesforce Unveils Agentforce 2.0. Wall Street Cheers Its AI Progress., Barron’s

Nvidia: Supplier-Led Blackwell Delays Can Push Out Production Timelines, Seeking Alpha

Apple working with Broadcom to develop AI chip, the Information reports, Reuters

Broadcom’s ‘Nvidia Moment’ Has Arrived. Now It Needs to Deliver, Bloomberg

Amazon ready to use its own AI chips, reduce its dependence on Nvidia, Ars Technica

Google and Samsung debut Android XR amid VR, smart glasses push, Yahoo! Finance

Harvard Is Releasing a Massive Free AI Training Dataset Funded by OpenAI and Microsoft, Wired

Chatbot answers are all made up. This new tool helps you figure out which ones to trust., MIT Technology Review

OpenAI Unveils New A.I. That Can ‘Reason’ Through Math and Science Problems, The New York Times

The AI war between Google and OpenAI has never been more heated, Ars Technica

Research: How Gen AI Is Already Impacting the Labor Market, Harvard Business Review

- Doug McCord

(Staff Writer)

Check out other articles from PTP on AI 

Get the latest updates on recruiting trends, job market, and IT, and expert advice on hiring and job seeking at The PTP Report