what to know
- AI can read your thoughts in a very real way by interpreting the signals in your brain and inferring your thoughts based on this information.
- It's a non-invasive process, meaning the AI works by reading brain scan data, rather than relying on surgically implanted electrodes.
- It relies on the same type of technology that powers ChatGPT, Google Gemini, and Copilot.
Artificial intelligence reads thoughts by interpreting brain signals. Early attempts at reading minds in this way used invasive methods to measure electrical impulses in the brain, and some succeeded in coming up with single words or even short sentences.
Recent examples leverage deep learning algorithms and large language models to interpret non-invasive brain scans and obtain more reliable results.
AI mind reading begins with a functional magnetic resonance imaging (fMRI) scan of a person's brain. The scanned data is fed into a specially trained artificial intelligence program that uses a large language model to interpret it, in essentially the same process as providing text prompts to ChatGPT.
Artificial intelligence analyzes the data and then determines what the person is thinking based on a large database of other scans.
Some mind-reading AI experiments have used stable diffusion and similar techniques to turn people's thoughts and even dreams into images. Other experiments transform a person's thoughts into text that reflects their thoughts.
AI is trained to read thoughts using deep learning algorithms and large language models.
How models like ChatGPT are trained, the AI does not provide reams of text, but rather brain scans and corresponding data showing what the person was thinking at the time of the scan.
One of the early successes of mind-reading AI was using a custom-trained version of GPT-1, an early version of the technology that powers ChatGPT.
In this case, volunteers lie in an MRI machine for several hours while listening to podcasts that create data for the training process. The researchers then used the brain scans and corresponding podcast audio to train GPT-1. Newer GPT models are based on this approach.
Once an AI is trained on brain scans and corresponding data, it can use that training to interpret other brain scans.
One of the most important uses of mind-reading AI is to help people suffering from paralysis and other illnesses.
While there have been some successes using technology like Bluetooth to bypass injured spines and control robotic limbs, they still rely on invasive procedures that directly hack into nerves to read signals.
Non-invasive mind-reading AI could control prosthetics and other devices through brain scans rather than surgically installed electrodes.
This could allow someone who has lost an arm or leg to control a prosthetic limb, or allow someone who is paralyzed to communicate their thoughts via text or speech synthesis.
AI can realistically imitate sounds, so in theory, someone paralyzed in an accident could even speak in their own voice if there were recordings to train the AI.
While there are many exciting possibilities for mind-reading AI, the technology's potential uses also bring with it a number of ethical considerations and privacy issues. Issues around consent, privacy and data security are all key issues here.
The issue of consent is particularly problematic because the ethical implementation of this technology requires informed consent, which may be difficult or impossible to obtain in some circumstances.
For example, the technology could allow paralyzed people to speak, but obtaining the initial consent needed to use the technology on such people could be difficult.
In other cases, such as when an employer requires employees to use the technology, the inherent power imbalance may also raise questions about consent.
Privacy and data security are also important areas of focus for mind-reading AI. Since this technology can read your thoughts, it can make your inner workings so much that even your innermost thoughts are no longer private.
This is a question for potential users to consider, but the issue of losing this level of privacy also raises huge concerns about data security.
Personal data obtained through data mining and other methods is already a major commodity, so it is likely that data obtained through this mind-reading technology will also be valuable and sought after.
Legislation may be required to provide additional layers of security for this highly private information, and any data breach could have extreme consequences.
Today, you use some form of artificial intelligence on your smartphone to sort through photos and find common items like trees, beaches, people, and more. If you want to give it a try, you can use Microsoft's Bing AI by signing up for an account.
The software works with the hardware to interpret your verbal commands and answer questions or requests. They're programmed for the most common uses (for example, "Hey Siri, add cheese to your shopping list"), and if they can't interpret the request, they send the audio file to a more powerful computer for interpretation. Once it knows your request, it consults the database and tells you the information.
