The beginning of the year is the traditional period of summaries and predictions for the coming one. It can certainly be said that last year belonged to generative artificial intelligence and large language models. A great many things happened not only in the area of its development, but especially in its popularization. I, on the other hand, am a person who prefers to look into the future, not the past, so I will focus today on predictions.
With such a tumultuous development of AI, any attempt at long-term prediction is more the domain of futurologists, so I will focus only on the coming year. And even in such a short period, I think it's more of a guess than a prediction, although some trends are also apparent. But what's the problem with taking up this kind of game? So don't hold it against me if I'm wrong. This is only a game.
I am aware that the audience of this forum is most interested in potential developments in the SAP and business areas, so the post will mainly refer to this field.
Companies vs AI
Undoubtedly, a niche has emerged in the market that many companies are trying to fill. On the one hand, we have the fight of leading companies in the race for dominance in the field of large language models with AGI on the horizon. On the other hand, there is also a lot of different types of startups developing dynamically innovative products, although not always useful in the classically understood business, at least not yet. In the middle between these two categories is a third, which includes a whole bunch of companies watching developments closely and considering how to use new products in optimizing business processes. While some of them have already been able to implement something, 2024 could see a rapid increase in spending on such projects.
Areas of AI development
There is no doubt that people will now create large language models with more and more parameters. Perhaps in 2024 we will see the successor to ChatGPT 4. On the other hand, the risk in business applications is the potentially greater tendency to hallucinate with more parameters. Therefore, the trend of making smaller but better-trained models, also with the help of their bigger brothers, will also be noticeable. It is already noticeable that smaller models are able to achieve results unimaginable just a few months ago. Smaller models also have a very high potential when it comes to their local application in devices, such as our smartphones. Hence, the battle for optimization in this field will be very visible. On the other hand, I am convinced that in 2024 we will not achieve or even come significantly close to the Holy Grail of this industry, which is Artificial General Intelligence.
In a separate category I would add the development of AI agents, performing more and more autonomous functions and also teams of agents working on more complex tasks. Perhaps the concept of agent teams is only a transitional form, while in 2024 we will hear about it more than once, including in the context of complex business tasks.
Multi-modality could also be a potential breakthrough. However, this will still be quite a challenge in 2024, even just because of keeping the balance between media. This has a very high potential, while at least it will not inhibit the development of models dedicated to one type of medium, at least for now.
AI-based robots, including humanoid ones, should be classified in a separate class. In 2024, they will certainly attract public attention. It will turn out that robots are able to learn very effectively by observing the actions of humans or other robots. They will certainly amaze us more than once. On the other hand, this year I don't expect it to have much impact on business yet, at least not in a mass sense. Let's remember that in narrowly specialized tasks, automation is already highly developed. It works great in factories or warehouses, and wherever work is highly repetitive. Of course, even there elements of artificial intelligence, especially vision AI, are able to improve quality. However, it is difficult to imagine that processes can be automated by introducing general-purpose robots instead of dedicated ones. Because what will be the advantage of future robots is general-purpose applications - where at the moment humans are difficult to replace. And although I expect clear progress in this area, not mass implementations.
In 2024, we will hear an incredible number of times that a device is AI-enabled, smart, intelligent, autonomous, cognitive, copilot or assistant. Admittedly, we've considered devices "smart" for a long time, but now it will take on new power. AI will be the two most popular letters in companies' public announcements. Some of these claims will even be justifiable. There will probably be at least a few AI assistants in the form of separate devices, while probably all new smartphones will be presented as AI-enabled. I expect that some of them will only connect to cloud-based assistants, but some may include built-in models. Processors or dedicated chips will probably be optimized for the inference process for local pre-trained models.
I think symptomatic for 2024 will be the increasing problems with access to the data required to train models. This is clearly shown by the New Your Times lawsuit against OpenAI and Microsoft, and the closing of APIs of websites such as Twitter and Reddit. I'm curious to see if this will also inhibit a bit the development of open source models, which are bravely chasing the quality of the leaders. A little in spite of everything, I hope that this will also be the year of open models. Therefore, another Holy Grail of this universe - synthetic data - will be even more welcome. Or will legislation reducing authors' rights solve the problem? Countries will be able to do a lot to have an advantage in AI field.
AI in service of business
In what areas can we expect the application of artificial intelligence in business to grow? An obvious application is data analysis. We are collecting more and more of it, and artificial intelligence is great for spotting trends, anomalies or preparing summaries. AI tools in the hands of an experienced data analyst can work miracles. Certainly, the productivity of experienced professionals will increase significantly, while reducing the need for less experienced employees. Just mention here Joule, who is an AI assistant in the SAP universe. I expect many more people to meet it in 2024.
The whole area of customer interactions, especially relationship management, is also a natural application. We will have better and better assistive tools, preparing ready answers to customers for employees. The ability to teach an assistant with specific knowledge along with great use of natural language will eventually open the way to helpful, not just annoyance-causing chat-bots. We can expect more and more customers to be convinced by such solutions and at the same time their quality will increase. SAP is also clearly anticipating this trend, as it recently announced new AI capabilities in its SAP CX (Customer Experience) product.
AI understands our needs better than we do ourselves. That's why I expect a tumultuous development of artificial intelligence applications in marketing. Already many companies are using, for example, tools to generate content, or specifically images, especially in the area of social media marketing. These applications will expand rapidly into other areas. Probably generative artificial intelligence is not yet able to replace copywriters, while, as with other specialists, it greatly increases the productivity of experienced employees. It can also verify what has a chance to go viral and what doesn't. I think 2024 will be an eventful year for the development of such tools and the increasing mastery of their application.
Meanwhile, I also expect some resistance from people and tiredness of content created by artificial intelligence, which will also be increasingly abused. The ease of generating distinctive, creative images will result in an oversaturation with them. The situation will be aggravated by automatically created articles that do not contain much content. In the distant future, this may lead to a greater appreciation of human labor, and perhaps even the creation of "AI-free zones". Especially since we will be surrounded by ever-present automated telemarketers, and scams using artificial intelligence will become more common. Which will also potentially result in hostility and fear from humans.
Applications closer to technology
I also imagine that artificial intelligence will have increasing applications in cyber-security. Here, I expect to see, first of all, the expansion of existing systems with intelligent functions. The question for SAP is to what extent Enterprise Threat Detection will be adapted to new trends, or whether the current real-time intelligence is already outstanding. It is also not excluded that something new will emerge. Unfortunately, one can also expect that the other side of the barricade will also use artificial intelligence to hunt down vulnerabilities.
Various types of copilots have been helping programmers not only to create code, but also to analyze it. Even if AI is not yet able to write entire programs on a larger scale, the productivity of programmers is certainly significantly higher thanks to new tools. This reduces the need for employees, especially at the junior level. And this is where development will certainly still take place. Programs will still be imperfect, but we are already seeing a trend and systems improving them in subsequent iterations based on error messages. I expect that in the future AI may self-correct its creation based on its own automated tests and perceived deviations from specifications. And since code creation is only a small part of the software development process, one can also expect the formation of teams of agents trying to carry out the entire process from gathering requirements, through technical project, coding and testing. I don't expect final results in 2024, but we will certainly have some attempts.
So far ABAP has been less subject to these trends, but last August we saw a demo in which an AI tool was able to generate the structure and code of a Fiori application. For now it's just a clear indication that SAP is working to make something available in this area, not a finished product, but who knows if we won't see a working prototype in 2024. I, however, see the opportunities elsewhere. Actually, for any consultant working with SAP, the ability to read ABAP code is a great asset. I expect that this is where AI can prove itself, explaining what code passages do. This would be helpful for many consultants of various modules.
In today's IT world, few things are more important than integrations. The era of standalone systems performing isolated functions is long gone. Everything interfaces with the world, often using external buses or platforms. I think it is precisely such services that have a chance for additional growth in 2024, if they offer, in addition to classic mechanisms, easy integration with large language models and easy construction of AI solutions. In my opinion, this is also an opportunity to popularize SAP BTP, which has built-in AI capabilities and also the ability to connect to various large language models. SAP BTP has the potential to seamlessly merge traditional SAP systems with the modern world.
Industries
Of course, it is important to realize that AI adoption will not be the same in all industries. This is nothing new, after all, there are more traditional fields, such as banking, where I expect even resistance from customers, should the custodian of the money come up with too many innovative management solutions. Of course, even here, for example, efficient fraud detection is welcome, but in general, caution should be exercised. At least in the publicly visible part.
However, there are industries where AI implementations can bring the most benefits - business-wise, but also in human terms. In general, it seems that industries that traditionally have ties and collaborations with the academic world have the greatest potential. I would particularly highlight healthcare, broadly defined, here. AI can help with drug development and diagnostics. What's interesting, by the way, even when interacting with patients, AI has an advantage - it can be more empathetic than humans. Sad, isn't it?
A great deal of good AI can do in education as well. This year I expect to see a great number of tools dedicated to supporting our learning. Providing a virtual teacher, tracking our progress, adjusting the level and content according to our progress and correcting mistakes meticulously. This is a very great potential for human development. Although, of course, in the long run, unfortunately, certain skills will tend to disappear in us when we are supported by AI assistants at every procces.
Retail can better understand customer behavior and adjust offerings to them through data analysis. In online commerce, we will have better and better hints of products we might want to buy. Analyzed not only on the basis of our previous purchases, but also our activities and plans. Not insignificant is also the development of marketing mentioned earlier, which will make advertising content to us more personalized.
The game industry had its golden time during the pandemic, so now it's expected to be a bit worse. Now we are going to face a very big optimization of content generation processes, leading to lower costs also through headcount reduction. On the other hand, thanks to AI agents, games should become even more immersive and attract more customers, but probably not yet this year. Although ultimately people should have more time, so maybe in the future we will devote much more time to this entertainment.
Summary
Yes, 2024 could be as interesting as the previous year. How the development will go, in my opinion, depends heavily on how efficient the next generation of LLMs will be. And also on how big a problem the tendency to hallucinate will be. Some hopes are currently being pinned on Mamba, a competing architecture to the Transformers, but it's hard to judge whether it has any chance of bringing something new. Or will something entirely new emerge? Although there doesn't seem to be much going on in this area.
We certainly live in interesting times. As I said, any predictions are just guesses. Therefore, we will watch closely how the history is made.
- SAP Managed Tags:
2 Comments
