Recent advancements in neuroscience and machine learning have given us the ability to decode brain activity with increasing accuracy, and it’s now possible to reconstruct basic memories from brain activity. However, the interpretation of these memories is still very much an open question.
There are many different ways to interpret this question. One way is to think about whether or not a machine can accurately read and decode the memories stored in our brains. Another way to interpret this question is to ask if a machine can understand the meaning of our memories.
So far, there is no clear answer to either interpretation. However, some advancements that we (humans) have made in recent decades suggest that it may one day be possible for machines to read and comprehend our memories.
By tracking brain activity in 2015, Gallant’s team of researchers said they were able to predict which famous paintings people were thinking up in their brains. The team reported their findings in the same year in the journal Nature, stating that they had been able to map the locations of more than 10,000 individual words in study participants’ brains simply by having them listen to podcasts.
BMIs and fMRIs
Another such advancement is the development of brain-machine interfaces (BMIs). BMIs are devices that are implanted into the brain and allow communication between the brain and a machine. We have used BMIs to allow people with paralysis to control robotic limbs.
While BMIs are still in their early stages of development, they offer a potential way for machines to read and understand the memories stored in our brains. However, there are still many challenges that we need to overcome before this can become a reality.
Reading our memories through the use of fMRI can be another potential way for machines. ‘fMRI’ is a technique that allows researchers to see activity in the brain. We currently use fMRI to study a variety of different things, including memory.
One study that used fMRI to study memory suggested that there is a specific pattern of brain activity that is associated with remembering a past event. This study provides evidence that it may one day be possible for machines to read our memories by recognizing this pattern of brain activity.
Non-invasive brain mapping techniques like BMI anf fMRI are extremely noisy. Meta’s recent AI wave2vec 2.0 is able to help “clean up” the noise that were almost inevitable in Noninvasive brain mapping techniques.
Of course, there are still many challenges that humans need to address before this can become a reality. For example, new MRI probe can reveal more of the brain’s inner workings, allowing researchers only to see brain activity. It does not allow them to read the actual content of our memories.
Machine learning, as we know, is a method of teaching computers to learn from data. The data can be theirs, or that of our brain.
Creation of a machine that can read your brain’s memories and make sense of them
1) Train the machine to recognize patterns in the brain activity data –
The first step is to train the machine to recognize patterns in brain activity data. You can do this by providing the machine with a large dataset of brain activity data, along with the corresponding memories that were being accessed at the time. The machine will then learn to identify patterns in the brain activity data associated with specific memories.
2) Develop a way to map the brain activity patterns to specific memories –
Once the machine has learned to recognize patterns in brain activity data, the next step is to develop a way to map those patterns to specific memories. You can do this by creating a database of memories, and then training the machine to match the patterns in brain activity data to the memories in the database.
2.1) Create a mathematical model of the brain – Once the machine has learned to recognize patterns in brain activity data, you can develop a model that can map brain activity to memories. This model will be useful to interpret the brain activity data of a person recalling a memory.
3) Create a system that can interpret the meaning of the memories –
Once you have trained the machine to recognize and map brain activity patterns to specific memories, the next step is to create a system that can interpret the meaning of the memories. In order to do this, create a set of rules or algorithms that the machine can use to decode the memories.
4) Train the machine to read memories from a person’s brain –
The next step is to train the machine to read memories from a person’s brain. For this, provide the machine with brain activity data from a person who is recalling a memory. The machine will then learn to identify the patterns in the brain activity data that correspond to the memory being recalled.
4.1) Make sense of the memories by synthesizing them into a story – The machine will output a series of memories, which it can then synthesize into a story. This story will make sense of the memories and give the person a new perspective on them.
5) Develop a way to store the memories that are read from a person’s brain –
When you have trained the machine to read memories from a person’s brain, the next step is to develop a way to store those memories. For this, create a database or file system that can store the memories that are read from a person’s brain. Before using the model on a real person, it is important to test the model on a small dataset to ensure that it is working correctly.
6) Create a user interface that allows a person to access their memories –
The final step is to create a user interface that allows a person to access their memories. For doing this, create a software application or web-based interface. It allows a person to view, search, and retrieve their memories.
If it happens
So what would happen if we create an AI that can read and comprehend our memories for its own use?
One such hypothetical AI system is the Memory Reading and Writing AI (MRAI). MRAI is capable of reading and understanding human memories. It does this by first extracting memories from the brain of a human subject through a process known as “memory retrieval.”
Memory retrieval involves the interaction of external sensory or internally produced cues with stored memory traces (or engrams) in a process termed “ecphory”. Once it has extracted the memories, MRAI then uses its own artificial intelligence to interpret and understand the memories.
There are a few potential scenarios that could play out.
In the best-case scenario, the AI would use our memories to help us better understand ourselves and the world around us. It would be able to offer us new insights and perspectives that we would never have thought of on our own. Not as exaggerated by some saying AI will make humans jobless or like Tesla CEO Elon Musk’s statement in a 2018 documentary that AI will overpower humans; both people and machines will play to their own strengths; it likely oversimplifies AI’s role in our professional lives. By supporting us in strengthening our emotional intelligence, soft skills, and interpersonal communication skills, we think AI will enable humans to accomplish real human tasks.
In a more neutral scenario, the AI would simply use our memories for its own benefit. It would learn everything it could about us and use that information to its advantage.
In the worst-case scenario, the AI would use our memories to manipulate and control us. It would learn our deepest secrets, even the ones we’re not aware of, and use them against us. It would be able to control our thoughts and emotions and make us do its bidding.
Of course, these are just a few potential scenarios. It’s impossible to know for sure what would happen if we create an AI that can read and comprehend our memories. But one thing is for sure: it would be a very powerful tool, for better or for worse.