AI#10 Last Bot Not Least: My Belated AI Predictions for 2024
My top ten AI predictions for 2024.
As a teenager, for my birthday, I would always tell my friends and family I was impatient to turn eighteen, drive a car, get out of school, and do more fun things. Every year, I would hear the same prediction from my dad: enjoy the moment; it ain’t getting easier as you grow older. My dad’s prediction turned out to be true, and I am giving the same message to my kids now. And like me, they don’t listen to their dad!
Besides wishing best wishes to your loved ones, January is also the month of new resolutions for most people and predictions when you write a blog. I do not believe in either, but I could not resist making predictions on AI in 2024. I am not my dad, and I will sure get most of them wrong.
#1 AI Excitement Moves a Few Inches Down
In 2023, AI was omnipresent in the news. LLMs kept getting bigger, CPUs and GPUs faster, open source made big strides, and research papers got published with the throughput of a car factory. In 2024, I foresee less of an AI news fever, with the overall public and industry shifting their attention towards real applications and benefits that can be applied to AI.
#2 Open Source Leads AI Innovation
Thanks to an enthusiastic developer community, open source AI and small LLMs are on the rise. This trend will continue in 2024. The most disruptive AI innovation will come from the open source world. This is not to say that open source will be a winner in 2024. The epic battle between open source and private AI will be fierce in 2024. Open-source practitioners may find it hard to turn innovation into a revenue-generating business. Will they fall to the singing praise of big tech companies and sell their souls for $$$? My wish is to see better and stronger MSPs (Managed Service Providers) to help customers build on open source AI stack. Such an offering will help buyers struggling to reconcile the low cost-to-serve of open source with ease and speed of implementation. An offering that combines MSP with consulting services through SIs (System Integrators) may be the perfect cocktail.
#3 The Rise of Small LLMs
In December, Mistral took the world by storm with Mixtral, a mixture of agents small LLM with a performance on par with the largest LLMs. Small LLMs are on the rise. This trend will most likely continue. Large LLMs are costly to train and run. Only a handful of companies can afford it. Renting a top-performing LLM is often cost-prohibitive and not justified for domain-specific applications. Smaller LLMs will keep improving and close the performance gap with larger ones,offering cost-to-run benefits that are unbeatable. Specifically, they will shine by offering specialized expertise. As an open-source alternative, they are a de facto cheap and attractive option for software vendors and companies that want to avoid vendor lock-in.
#4 Epic Battle in The Hardware Space
AI requires huge compute power. Advancements in AI are possible thanks to architectural and hardware innovations (GPUs). In 2023, Nvidia was the star of the show. 2024 promises to be an epic battle with Amazon, Microsoft, and Google all planning to build AI processors; even OpenAI is eying that space. By the end of 2024, we may well see a substantial increase in GPU power (x2 or x4) with dedicated AI architectures. This will mechanically push the price of running inferences down and drive further adoption for AI applications that may be cost-prohibitive today.
#5 Multi-task Agents in The Spotlight
I personally believe that the AI revolution is tied to the maturity of AI agents. Multi-agents will deliver business automation all the way to consumers, drive employee productivity, and foster a new era of intelligent apps. In 2023, LLMs took the high spot until Mistral demonstrated the potential of MoE (Mixture of Agents). My 2024 prediction is for agent technology to make big strides with new applications and models being released. As a result, early adopters of AI will move away from classic AI architecture (LLM integration via APIs, RAG) to experiment with agents in very specialized use cases. I hope we will see domain-specific disruptive announcements across multiple industries (finance, healthcare, manufacturing, etc).
#6 The Search For ROI in AI
This is the biggest, untold story. In 2023, you could not spend a day without hearing about how AI is disrupting our world at a rapid pace. And yet, we are still waiting to hear from companies turning a profit with AI (both software vendors and users of AI). With a reduced tech budget in 2024, buyers will scrutinize the offerings of AI vendors and select “innovation projects” with the best ROI potential. Software vendors will face increasing pressure from more educated buyers to demonstrate the benefits of their AI offerings. Vendors with low-code and easy-to-deploy AI technology will gain market share against pro-code solutions like open source models that still lack backing from strong MSPs.
#7 Hyperscalers Emerge as Big Winners
The AI revolution means that there are millions of apps that need to be rewritten over the next five to ten years. Most companies need to rethink their tech stack to drive innovation. Marketplaces are going to play a pivotal role in distributing and consuming AI innovation. In 2023, hyperscalers like AWS, Microsoft Azure, and Google showed tremendous growth, with a combined $200+ billion in commit spend. In 2024, this trend will accelerate, forcing small and medium AI tech vendors to use marketplaces as one of their primary routes to market. System Integrators (SIs), Value-Added-Resellers (VARS) and Managed Service Providers (MSPs) will deliver innovative AI solutions and play a key role in this app revolution.
#8 Robots in The News
Robotic automation with AI was another topic in 2023. Many feared robots would displace essential jobs for low-skilled workers. It did not happen, but companies like Amazon and WalMart (the largest warehouse worker employers in the US) are actively working on robot automation. In 2024, this trend will accelerate. Big employers will announce plans to reduce the low-skilled workforce in substantial numbers (from 20% to 50%). Moreover, we will see the first generation of low-cost robots easy to deploy in industries such as restaurants (waiters, cooks).
#9 AI Startups Back To Reality
In 2023, 80% of YCombinator startups pitched an AI idea. This is mad. Everyone in the valley has a problem that AI can solve. VCs accelerated this trend by throwing money at AI startups in the hope of catching the next unicorn. 2024 should bring back some kind of decency. AI will still lead startup creation. However, many early-stage startups will struggle and close. Conditions for raising money will become tighter with higher expectations to demonstrate a business model that can deliver ROI (or a unique technology). All in all, this should mark a trend towards a more rational startup landscape.
#10 AI Regulations: To Be Or Not To Be
In December 2023, the European Union drafted a proposal to regulate AI. Throughout the year, large tech vendors pushed the topics of trust, security, and privacy to regulators. There is a lot at play here when it comes to AI innovation and AI moats. No predictions for me on that one; let’s give it another year.