At the start of 2023, it would have been appropriate to describe the general prospects of the tech sector as bleak. High-interest rates, a crypto winter, and layoffs across the industry pointed to a tepid year. Nearly 12 months later, prospects are looking up thanks in large part to a single phrase now heard around the globe: artificial intelligence.
Recommended AI ML Posts: 10 AI ML In IT Security Trends To Look Out For In 2024
NVIDIA – the leading AI-powered company on the NASDAQ – is trading well over 200% above where it started in January of 2023. Microsoft and Meta are up 50% and 200% respectively, all on the promise and inclusion of AI capabilities into their products. At the center of this market enthusiasm is a singular functional presentation, made popular by OpenAI’s runaway success with ChatGPT: bots.
The appeal of chat-based AI is unsurprising. A good conversation is a familiar way to assess and benefit from intelligence. Alan Turing’s eponymous test codified this as the goalpost for AI in 1950, and Stanley Kubrick embodied AI in the popular imagination as Hal – a powerful (and dangerous) disembodied voice. Suddenly, after years of repeating instructions to early AIs like Siri and Alexa, we finally have a bot worth talking to. ChatGPT has emerged as the fastest-growing product ever by being able to actually help with research, writing, and more.
Bots have become a way to put a face on the AI revolution.
Fears center around rogue AIs who are able to outwit humans to cause harm, or around how a research bot might help a terrorist build a virus or bomb. Meanwhile, businesses rush to deploy “agents” to replace representatives or “copilots” to assist engineers.
But the familiar packaging and function carry forward human limitations. Bots are reactive, responding only to questions that we know, and have time to ask. Their ability to directly interact with data is bounded, requiring external curation and bolt-on tools such as search to bring the right information into focus. And their quantitative reasoning is, for now, abysmal. Like us, bots are suitable for some work – but not all. But if we look under the hood, the incredible advances in the technology that power bots can do so much more than simply field and return a response.
AI Trends in 2024: Thinking Outside the Bot
2024 will be the year that AI advances outside the familiar and relatable framework of “call and response” bots. Outside the commercial spotlight on OpenAI and Google’s most broadly capable (and expensive) foundational models, a thriving open research community has been advancing large language models (LLM) that can be applied cost-effectively, at extreme scale, for specialized applications. Open-use foundational models such as Meta’s Llama2 and Mistral’s MPT, paired with new specialization and cost reduction techniques such as LoRa from researchers at Microsoft and Carnegie Mellon University, are changing the landscape of what developers can implement without sending data to an external API.
These models will lend new AI capabilities to data analysis on a massive scale, and produce proactive features that can work on our behalf, alerting us to what is important before we know what questions to ask. Consider an “always on” AI that can flag potential discrepancies in medical patient charts, or a contact center app that automatically detects and escalates logistical challenges, or websites that effortlessly micro-personalize the most relevant results for individual users.
As other forms of generative AI steal away mindshare from chat-based AI, regulators in particular will begin to realize that current frameworks for AI management and harm reduction may be insufficient for the task. For example, the Biden-Harris Administration’s recent executive order emphasizes “red teaming”– prompting an AI to break its own rules – at the level of foundational models. For non-chat systems, more risks may emerge from the way an AI is deployed in the context of other technologies.
Then there are the executives who lead commerce. Many are now, or have already been building plans that treat the AI opportunity as either:
1) a bot for every customer,
2) a bot for every employee, or
3) both.
When these executives begin to see competitors applying new forms of AI in other ways, they will be forced to revisit commitments to vendors and policies that are mostly, if not entirely, bot-focused.
And herein lies the place where this past 12 months may show itself to have been a strong growing pain for AI, but not yet a mature sector.
The full capabilities of AI are only just beginning to reveal themselves. We are, quite literally, only in the opening chapters of what AI will ultimately become.
But in a rush to decide, or in a rush to secure greater business efficiencies, or in a rush to comprehensively legislate, we may find ourselves in a game of Monopoly being sent back to Go.
These new technologies will offer new challenges in how to establish consumer protections, in how to organize businesses to take maximum advantage, and how to adapt as users to a changing world. But it will also spread the benefits of AI beyond the usual tech giants, into a broader range of businesses and roles, a rising tide to lift all boats.
Top Story of the Month: Optimizing AI Workloads and Storage: From Data Collection to Deployment
Comments are closed.