9 Trends in Artificial Intelligence

?? This is a?signal?from the future.?More than 14.000 people receive this signal. You can use it?to reflect on yourself, society or your business.

Thanks so much for reading.

Kind regards, Jarno Duursma.

Artificial intelligence | Author | Speaker

Please connect with me on LinkedIn!

Contact? [email protected] and www.jarnoduursma.nl

Hi friends,

It's my job to closely monitor developments in the field of artificial intelligence, but occasionally also to zoom out and look at the bigger picture. These are the nine trends in AI that you should keep an eye on in the coming months.

Do you have any additions? Let me know in the comments!

001: In-house customized Language Models

First, larger companies are actively working on customizing a large language model. (Being open source or closed). They do this not only from a perspective of privacy and security but also because they realize that owning a language model can provide a huge competitive advantage. A customized language model, trained on all their digitized documents such as contracts, tenders, emails, and grant applications is an unprecedented source of new knowledge and information. It allows you to gain insights, analyze trends, and distill knowledge not previously known.

Through retrieval augmented generation (RAG) technology, you can add your own information and increase the reliability and currency of your answers. Keeping your training and RAG in-house also prevents your data from having to go through third parties, which is good for your privacy and security.

002: Open-source Language Models

The second AI-trend is related to the first. Open-source language models are getting better, such as Mistral Large and DBRX. They are approaching GPT-4 quality. They also offer organizations the opportunity to develop their own custom AI models relatively quickly – trained on their company-specific data and tailored to their specific needs – without exorbitant investments in infrastructure.

(Open source) models are also often getting smaller, meaning they will eventually be available on every device, primarily laptops and smartphones. Every device will have its own AI-assistant, which can also be used without an internet connection.

003: AI Productivity Teams

The third AI-trend is creation of "AI productivity teams" within companies. Management or board give mandate to these teams that are exploring how the new wave of AI tools can offer an advantage for their specific business processes and documents, and how they can adapt the toolkit of possibilities to their own company.

004: Increased Processing Capacity of Language Models

The fourth trend is about the increasing capacity of large language models to process more data. Where you now sometimes get an error message when entering a lot of text, that will eventually be something of the past. For example, Google's new Gemini model can handle 10 million tokens, which equates to 17,000 pages. This makes it increasingly normal to be able to give large amounts of documents to a language model to distill knowledge, discover connections, or generate new insights.

005: Text to Video

The fifth trend is the development from text to video, with which you can create a full high-definition video with a single sentence or a short description. We've all seen SORA. Visual ideas can now be generated faster than ever, a development we already know from images, but this year high quality text-to-video will also become available. This will lead to an explosion of new forms of film and video, made available to everyone, .. and thus the risk of disinformation also increases.

006: Multimodality in AI

The sixth trend is the development of multimodal systems, where AI systems can easily handle different types of digital content like audio, text, video, computer code, etc. As a result, it ultimately makes no difference whether a system is fed with text, video, or both. Smart systems can just as easily distill information from text as from video or audio. The possibilities will be enormous. Gemini 1.5 Pro can process vast amounts of information in one go — including 1 hour of video, 11 hours of audio, codebases with over 30,000 lines of code or over 700,000 words.

Particularly from video, a tremendous amount of information can be distilled.

007: AI for for scientific breakthroughs

The seventh AI-trend is that AI can accelerate scientific research. AI is a really good tool for finding patterns in data, formulating theories and testing them, analyzing models more quickly and searching through scientific papers more efficiently. AI is a powerful for scientific breakthroughs. (one, two, three)

In the future, AI will also be capable of explaining logical steps in reasoning. Contrary to what the general public thinks, this is not yet the case. Language models can indeed show the correct answer to a question, but why they generated that answer is often difficult to explain.

We will see spectacular developments in scientific research, when intelligent software can create, test, and reason hypotheses. Although it will be a while before we get there, we are already seeing the beginning.

008: The Autonomous Software Agent

The eighth AI-trend is that of the software agent. This is not only about the digital butler, about which I already wrote in my book in 2017, but the next step is the autonomous software agent. An agent that autonomously performs tasks on your laptop, smartphone, or computer system. OpenAI is working on it. You give a command, and the system divides it into small tasks to then execute them.

"Those kinds of requests would trigger the agent to perform the clicks, cursor movements, text typing and other actions humans take as they work with different apps" (OpenAI)

I don't know when a software system will be able to perform any conceivable laptop task at a level comparable to that of the average human.

009: Ethical Use of AI

Besides focusing on what AI can do for companies, there is a growing awareness around AI ethics. Employees are increasingly aware of the ethical implications of their AI products and are asking questions within the companies where they work. Think about bias, energy consumption, or mass surveillance. Consumer demand for ethical AI is also increasing. They are becoming more value-driven and increasingly demand ethical accountability from producers of such tools.

Thanks for reading the SFFT newsletter! Ask your friends to subscribe. :-)

Kind regards, Jarno