These five AI trends will change our lives in the years to come - TMC (nl) Shape caret-double-left caret-double-right caret-down caret-left caret-right-circle caret-right Shape close dropdown expand more facebook Logo linkedin logo-footer logo-mark logo-mobile mail play search twitter youtube instagram
Menu Sluiten

These five AI trends will change our lives in the years to come

Texts that are written automatically and unique images that are generated in seconds... In recent months, artificial intelligence (AI) has shown just how much it is capable of. If developments continue moving this rapidly, a lot could change in the years to come. We have listed five AI trends to look out for.

1. Stable Diffusion: creating new images at lightning speed

The ability to generate an image in seconds with just a few keywords using tools like DALL-E and MidJourney is every bit as impressive as the much-discussed ChatGPT. Like ChatGPT’s language model, these models make use of the massive amount of data available on the internet, in this case existing images.

Stable Diffusion is also based on a text-to-image model. The difference with the other two is that it is an open source model that anyone can run by themselves on a somewhat powerful computer. Stable Diffusion uses deep learning. The model is trained to remove noise from an image. It first receives an image full of noise, and asks it to generate an image with slightly less noise. It is able to create a sharp image by repeating that process. Based on the information the model has about what an image should look like, it slowly starts adding structure to the noise.
These types of tools are rapidly being developed and fed with more and more data. We will see plenty of other examples of generative AI in the years to come that will be used to generate content such as entire videos or pieces of music based on descriptions.

2. Diagnoses: better than doctors

Deep learning models that recognize images are not only useful for generating new images. They have also played a role in healthcare for years, helping doctors make diagnoses. Several years ago, for example, scientists developed a model that could recognize skin cancer from photos of patients better than dermatologists could, regardless of how much knowledge and experience they had. This kind of neural network is trained using tens of thousands of images. For each image, it can assess whether it is cancerous or not with an accuracy of up to 95 percent.

Is that bad news for medical specialists? Absolutely not. Although these models can be used to help doctors make diagnoses, doctors still have the final say. If the models work well, they also save doctors a lot of time that can then be spent on research or other important work.

3. Mind reading: is it possible and is it allowed?

Perhaps the most profound - and terrifying - possibility with AI is the ability to read someone’s thoughts. The first question that arises is whether this is even possible. CEO Mark Zuckerberg of Meta (Facebook) envisions a future where people can send thoughts directly to each other if they want. While it is highly doubtful that it will ever get that far, there is plenty of research being done on brain-to-brain communication and there are actually systems in existence that are able to take the first steps. For example, the research team from Braingate is investigating how a brain/machine interface that uses electrodes can enable people with brain disorders to control robotic arms and tablets with their thoughts.
Although that may very well be a valuable application, ‘mind reading’ could also lead to a dystopian future, as a video from the World Economic Forum shows. In this video, the brainwaves of a company’s employees were decoded to determine how seriously they take their jobs, whether they have certain feelings toward their co-workers and whether they are involved in illegal activities. Electrodes in the brain are not needed to record all that information. It is enough to have a wearable such as a smartwatch or earbuds.

For now, this kind of application can only be seen in series such as Person of Interest. But as one researcher pointed out, what you think and feel can be reduced to data. And what if AI does get extremely good at interpreting large amounts of data. Ethical issues and privacy concerns would probably hinder that development more than technology.

4. Entity resolution: combining data intelligently

Combining data is one of the most important tasks AI is capable of. But data are collected and stored in a number of different ways. To traditional computer programs, two sets of data are considered unequal if they are not 100 percent the same. With entity resolution, it is possible to link data that are nearly the same but not completely identical. It’s a clever way of discovering that highly similar data is linked to the same person.

Entity resolution can combine data in such a way that it can be traced back to a person. This is useful for companies that do not always keep data in their systems in a consistent manner, but allows the data to be used to improve patient records in order to deliver better care, or for fraud prevention. Again, the discussion of ethics and privacy is never far away. How far can you go in combining data? What happens if the algorithm was off the mark? How anonymous can we still be in our lives?

5. Singularity: will we soon become obsolete?

Will algorithms become so smart in the future that humans will be made obsolete? Some people claim that this could become a reality within a decade. The idea is not new – think of films like The Terminator. The reason behind some experts believing it will not remain science fiction is largely due to the speed at which AI and other technology is developing. In many cases, this is not linear development, but exponential. The question then is whether and when the tipping point will be reached where AI becomes so intelligent that it can improve its own algorithms without us humans having to think about it. To make people aware of the opportunities and dangers that technology holds and the speed at which it is advancing, the Singularity Group was formed and organizes regular events.

Will humans become obsolete as a result of AI?

With all of these AI applications, which are also developing at lightning speed, you might almost start to think that we could eventually be able to leave just about all of our work to algorithms. If you imagine an extreme form of singularity, then is it possible for all of humanity to even be replaced? Most experts don’t believe it will go that far. However, they do think that people who do not work with AI will be replaced. In other words, following AI with a critical eye but embracing it and collaborating with it in a constructive way seems to be the solution for the future.

Curious about data science at TMC? Reach out to us:

Bram Thelen

Director Data Science | Nanotechnology | Physics, Netherlands

Tel: +31 (0)6 52 89 25 70

Nieuwste van TMC

Stel je vraag