AI predictions for 2024
Dr. Salvador Salazar-Albornoz
Head of Data & AI | Mentor | working on Data Engineering, Science, Analytics and Machine Learning
In 2023, big strides in generative AI amazed everyone and put Artificial Intelligence right in the spotlight of everyone's talks and company roadmaps. Very few could have predicted the changes this brought – I certainly could not – and it seems like 2024 won't be any different. Still, I want to talk about a couple of things from 2023 that I think will stick around next year.
Non-Gen AI Dominance
Whether it's in healthcare, finance, or manufacturing, supervised learning is and will remain the largest driver of AI's impact.?
Even with all the buzz about generative AI, most of the value generated by AI applications does still come from more "traditional" types of machine learning. Whether it's in healthcare, finance, or manufacturing, supervised learning is and will remain the largest driver of AI's impact.?
In this regard, the real differentiating factor in this competitive scene is having good data. It's not just about having data; it's about having the right data, in the right amounts. In this data-driven age, success depends on this – like many say: "data is the moat".
Evolution of GenAI: ChatGPT and Beyond
Shifting gears to generative AI (GenAI), last year was a big deal with ChatGPT and other open-source Large Language Models (LLMs). These tools, once only for experts, are now out in the open, changing how people see AI. Regular folks are getting a better handle on what these tools can do, making AI a mainstream topic.
The biggest deal with LLMs though is not their eloquence but how they've transformed how we interact with computers. Over time we’ll see filling out forms and clicking buttons as something of the past. Now, we've got a new way of talking to technology that's more natural and easy for everyone – This is the biggest revolution.
领英推荐
The biggest deal with LLMs though is not their eloquence but how they've transformed how we interact with computers
Anticipating Multi-modal Generative Models
Looking at the next year in AI, all eyes are on multi-modal generative models. These models are like the next step after LLMs and are set to be the new big thing, changing how we use and interact with computers all over again. The exact details of this change are still unclear and hard to anticipate for most of us.
No AGI yet
Without downplaying the incredible technological revolution we are living through, during the last year we’ve also better understood the limitations of these quite eloquent language models. Understanding that these models are really “just” doing next-token prediction auto-regressively makes it a bit clearer why they fail where they do. Recognizing this is key to using these tools the best way possible without getting caught up in unrealistic dreams.
Conclusion
Nobody really knows what the future holds, but looking back at 2023, the surprises and changes brought by generative AI show how tech can shake things up. But there's still a lot we don't know about that. We've figured out that even the smartest language models have limits, so it's smart to keep our expectations real.?
As we head into the next year, the only thing we can be sure of is that the journey with AI will keep going. We can't predict everything, but learning from the past year reminds us to keep our eyes open and be ready to adapt.
HR Operations | Implementation of HRIS systems & Employee Onboarding | HR Policies | Exit Interviews
7 个月Great perspective. "The term ""algorithm"" is derived from the last name of Persian mathematician al-Khwarizmi, who presented the first systematic technique for solving equations. Traditional algorithms are well-defined processes or sets of rules for solving problems. Indeed, these algorithms are fixed and do not change over time or after processing more data. On the other hand, just like humans, Machine Learning algorithms learn and modify themselves as they process more data. Hence, in 1950s, the paradigm of traditional algorithms was upended by that of Machine Learning algorithms, and in Thomas Kuhn’s terminology, a scientific revolution occurred. Today, Machine Learning is a vast field that includes supervised learning, unsupervised learning, reinforcement learning, and mixed learning. Supervised Machine Learning involves humans training a computer program to classify data based on pre-labeled examples. Unsupervised Machine Learning techniques do not require pre-labeled data or a human trainer. Reinforcement Learning algorithms learn from the consequences of their actions and improve their performance through trial and error. Finally, Mixed Learning combines all these techniques.
Innovation | MBA | Digital Transformation | Sustainability | Thought Leader | Strategy | Coaching | LinkedIn Top Voice | Vision
10 个月Nice article, Salvador Salazar-Albornoz. I am looking forward to seeing the impact of Gen-AI on improving society in fields such as medicine and manufacturing.