Beyond Speech and Writing: Can AI Let Humans Exchange Thoughts at High Speed?

A scientific study of artificial intelligence, brain-computer interfaces and direct thought exchange between people

Scientific review article and proposed research framework | November 2026

Abstract

Today's modern computers exchange information at speeds that are hard to imagine.

An ordinary Gigabit Ethernet connection can transfer data at one billion bits (1,000,000,000 bits) per second. Advanced computing and data-centre technologies operate many times faster still.

But the human situation is entirely different.

Even today, people mostly use speech, writing, gestures and body movements to convey their thoughts, ideas, opinions and feelings to others.

Although enormous neural information processing takes place in the human brain, our capacity to transfer that information to another person in a meaningful way is comparatively very limited.

This situation raises an important scientific question.

Using artificial intelligence and neural technologies, can people exchange their thoughts and ideas directly, accurately and at high speed, without the help of speech or writing?

This article examines the speed of human information transfer, technologies that interpret the brain's neural signals, brain-computer interfaces, brain-to-brain communication and the scientific progress made up to November 2026.

It also proposes a new research framework for the high-speed exchange of human thoughts and intentions with the help of artificial intelligence.

The fundamental aim of this research is not simply to transfer more digital data.

Its central purpose is to explore the scientific feasibility of a new communication method that conveys to others the true meaning of what a person wishes to express, in less time and with greater accuracy.

1. Introduction: The Speed Barriers in Human Communication

Consider two computers connected to a modern computer network.

They can convert very large amounts of information into electrical signals and exchange them within a few milliseconds.

Now consider two people exchanging a complex idea.

The first person must form an idea in their mind.

Then they must convert that idea into language.

They must choose suitable words.

They must arrange those words into a proper sentence.

Only then can they speak or write.

Likewise, the second person must hear or read those words. They must then understand their meaning and work out what the first person actually wanted to say.

Every stage of this process takes time.

On some occasions misunderstandings can arise. The idea then has to be explained again.

The important problem here is not that the human brain works slowly.

Rather, there is a significant gap between the neural information processing that takes place in the human brain and the speed at which that information can be conveyed to others.

Research by Coupé and colleagues, published in the journal Science Advances in 2019, studied 17 languages and estimated that the information-transfer rate of human spoken languages averages about 39 bits per second [1].

In another important study, Zheng and Meister estimated the rate of human behavioural information processing at roughly 10 bits per second. That study was published online in 2024 and appeared in the journal Neuron in 2025 [2].

These two studies measure different kinds of information processing. The two numbers therefore cannot be used interchangeably.

Nor are they final measurements of the human brain's total neural processing capacity.

Nevertheless, they show that the information rate of conscious human actions and communication is far lower than the data-transfer rate of many electronic systems.

This difference forms the basis of our new research.

2. The Data-Transfer Speed of Computers and the Communication Speed of Humans

Computers convert information into digital form and exchange it according to defined communication protocols.

For example:

• 1 gigabit per second (1 Gbps) = 1,000,000,000 bits per second.

• 10 Gbps = 10 billion bits per second.

• 100 Gbps = 100 billion bits per second.

• 800 Gbps = 800 billion bits per second.

These figures describe the information-transfer capacity of a computer network.

But they say nothing about how quickly a computer understands the true meaning of the information being sent.

Human communication is entirely different.

In exchanges between people, what matters most is not only the amount of information but its meaning.

For example, one person may express a very complex decision clearly in a few words.

Another person may need several paragraphs to explain the same idea.

Great scientific care is therefore needed when comparing the data-transfer speed of computers with the communication speed of humans.

A computer may transfer a billion bits per second.

But that does not mean it understands the meaning of that information.

At the same time, a person may convey a profound idea to someone else using only a few words.

The aim of our research is therefore not to turn the human brain into a gigabit computer network.

Rather, our aim is to increase the speed at which meaningful thoughts and ideas can be exchanged between people in a way that is correctly understood.

This is the most important scientific foundation of this research.

3. The Progress Science Has Made So Far

3.1 Brain-Computer Interfaces

A brain-computer interface, or BCI for short, is a technology that creates a connection between the neural activity of the human brain and an external computer system.

Using this technology, signals obtained from the brain can be analysed by computer.

There are two main approaches to recording brain activity.

First, methods such as EEG that record the brain's electrical activity without surgery.

Second, methods that use electrodes implanted in the brain through surgery.

Artificial intelligence models analyse these neural signals and attempt to identify the action a person intends to perform, the word they want to express, or a communication command.

Research by Willett and colleagues, published in the journal Nature in August 2023, was an important advance in this area.

That work created a technology that converts into text the neural activity that occurs when a person attempts to speak.

The participant was able to communicate at a rate of about 62 words per minute.

The word error rate was 9.1% with a restricted vocabulary of 50 words and 23.8% with a large vocabulary of 125,000 words [3].

This research proved that brain signals produced when a person tries to speak can be used to communicate relatively quickly.

But this system is not a technology that can freely read every thought arising in a person's mind.

It identified the neural activity associated with the attempt to speak.

3.2 Generating Speech Directly from Brain Signals

Further notable advances in this field occurred in 2025.

Littlejohn and colleagues created a technology that processes signals from the human brain in 80-millisecond intervals and converts them into speech with the help of artificial intelligence [4].

In another study, Wairagkar and colleagues described a method that uses the brain's neural signals to produce expressive synthetic speech in real time.

The participant in that study was also able to control, to a certain extent, expressive elements such as voice intonation [5].

This is a very important advance.

Human communication is not merely the exchange of words.

The speed at which we speak, the tone of voice, emotional expression and the way we stress words all reveal the true meaning of an idea.

These studies demonstrate the feasibility of reconstructing natural human speech from neural signals.

Nevertheless, they remain experimental technologies used in specific research settings.

3.3 Long-Term Practical Use: A Landmark Study of 2026

A study published in the journal Nature Medicine in June 2026 showed an important advance in the practical use of brain-computer interface technology.

In this study by Card and colleagues, a person with severe motor paralysis communicated independently for a long period with the help of an implanted interface.

Over nearly two years, he used the technology at home for more than 3,800 hours.

Through it he communicated about 1.96 million words and 183,060 sentences.

The average communication speed was about 56 words per minute.

In controlled speech tests, word accuracy exceeded 99%. But in everyday natural conversations this accuracy was not the same at all times [6].

This difference is very important.

It is one achievement for a technology to work well in a laboratory.

It is another achievement for the same technology to work reliably for a long period in people's daily lives.

In our high-speed human thought-exchange research, both of these aspects must also be proven separately.

3.4 Can Information Be Exchanged Directly Between Human Brains?

There have also been some early successes in the scientific world concerning brain-to-brain communication.

A study called BrainNet was published in 2019.

In that study, three people worked together on a simple game using brain signals.

The decisions of the first two participants were recorded using EEG technology.

That information was sent to the third participant as defined signals through a brain-stimulation method called Transcranial Magnetic Stimulation (TMS).

Five groups took part in this experiment.

Their average performance accuracy was recorded as 81.25% [7].

This study demonstrated the feasibility of transferring certain information between human brains through neural interfaces.

But what was transferred was only very simple decisions.

Complex thoughts, memories or complete ideas were not transferred directly.

There is therefore a very large scientific difference between transferring a simple binary piece of information and transferring a person's complete thought to someone else.

3.5 Detecting Speech Imagined in the Mind

A study by Kunz and colleagues published in the journal Cell in August 2025 showed that the neural activity that occurs when a person imagines speaking in their mind can be partially identified.

Another important aspect of this research is mental privacy.

The risks of a person's private inner speech, which they do not wish to share with others, being accidentally revealed through technology were also examined.

In addition, safety methods were tested to distinguish between thoughts deliberately created for communication and private mental activity [8].

Another study, published in the journal Nature Neuroscience in September 2026, examined some common structures in the neural activity of imagined speech and overt speech [9].

These findings may create opportunities in the future for people to communicate certain information without even attempting to speak.

But there is no scientific evidence to say that current technologies can accurately understand every thought that arises in any person's mind, without prior training or cooperation.

4. The Proposed New Research Concept

High-Speed Human Thought Exchange with the Help of Artificial Intelligence

Artificial Intelligence-Assisted High-Speed Thought Communication

The central aim of this research:

To explore the feasibility of a technology that identifies, from the brain's neural signals, the thoughts and ideas that people voluntarily wish to communicate, converts them into an efficient digital form, transfers them to another person, and enables that person to understand the true meaning of the idea accurately.

This research brings together several scientific fields.

Brain-Computer Interfaces: used to record and analyse signals obtained from the human brain.

Artificial Intelligence: used for tasks such as detecting patterns in neural signals, language generation, understanding ideas and correcting errors.

Information Theory: helps to assess mathematically how efficiently and accurately information can be transferred.

Secure Digital Networking: helps to send the generated information safely to another computer system.

Brain stimulation and neural technologies: may help explore the future possibility of delivering specific information directly to the recipient's neural system. Extensive medical and scientific testing is essential for this.

The important opportunity in this research may lie in creating a method that conveys the meaning of what a person wishes to express, without needing to generate and transfer every single word they want to say.

For example, suppose a person thinks the following:

"I agree with this plan. But I would like to postpone implementing it until next week."

A future technology could attempt to identify the three important elements in this thought.

First, agreement with the plan.

Second, reluctance to implement it immediately.

Third, a request to postpone it to next week.

Artificial intelligence could identify these basic meanings and construct a sentence that the other person can understand.

But there is an important challenge here.

It is not enough that the sentence artificial intelligence generates is linguistically elegant.

It must express, unchanged, the idea the first person actually wanted to convey.

This is the importance of the concept called Semantic Communication.

5. The Proposed Technical Framework

The initial structure for implementing this research can be divided into six main stages.

Stage 1: Recording Brain Signals

A person decides, with their full consent, to convey information to someone else.

At that moment, the brain activity associated with the word, command or idea they are deliberately trying to express is recorded with the help of a suitable neural interface.

In early-stage research, non-surgical methods such as EEG and already published research data can be used.

Stage 2: Identifying the Thought with Artificial Intelligence

The recorded brain signals are analysed by machine-learning models.

The aim is to identify the information the person deliberately tried to convey.

Research should begin by identifying a small number of words, commands or ideas.

Only after the accuracy of that basic technology is confirmed should work move on to understanding complex thoughts.

Stage 3: Converting the Thought into Meaning-Based Data

The identified information is converted, with the help of artificial intelligence, into a structured digital form.

This can include the thought's main ideas, the relationships between them, conditions and the uncertainty of the estimate.

This makes it possible to represent the basic meaning of the information in a computer system even before a complete sentence is generated.

Stage 4: Secure High-Speed Information Transfer

The generated data is sent to another system through a secure, encrypted digital communication channel.

At this stage, the speed of the computer network need not be a major barrier.

Today's computer networks are already capable of transferring information at very high speed.

Instead, accurately understanding brain signals, generating the correct idea and reducing the time the whole process takes may be the more difficult problems.

Stage 5: Reconstructing the Message for the Recipient

Artificial intelligence converts the received information into a form another person can understand.

It may be text, speech or another suitable form of communication.

The recipient must be enabled to understand the meaning of the information, to ask for clarification if in doubt, and to know the uncertainty in the artificial intelligence's estimate.

Stage 6: Confirming That the Message Was Understood

It must be confirmed that the recipient has understood the idea correctly.

Where needed, the facility to request clarification or correction must also be provided.

If direct brain-to-brain communication is to be created in the future, a separate technology capable of delivering information to the recipient's neural system must also be developed.

An important distinction must be noted here.

Converting one person's brain signal into text by computer and having another person read it on a screen is brain-computer-human communication.

It cannot be called complete direct brain-to-brain communication.

This distinction must be maintained clearly at every stage of the research.

6. The Central Scientific Hypothesis of This Research

The central hypothesis for this research is proposed as follows.

By identifying, with the help of artificial intelligence, the neural information that people deliberately wish to communicate, and converting its meaning into an efficient digital form, certain ideas can be conveyed to others in less time than with speech or writing, while maintaining acceptable accuracy.

This hypothesis can be tested scientifically.

For this, the same kind of information can be exchanged through ordinary speech, writing and neural interfaces connected to artificial intelligence.

Wherever possible, the same participants and the same information should be used to compare the results.

The following measurements are important in this study.

1. End-to-End Latency

The time from when one person begins to convey information until another person correctly receives it.

2. Effective Information-Transfer Rate

A measure of how much information was correctly transferred per second, after accounting for errors and corrections.

3. Semantic Fidelity

The similarity of meaning between the idea the first person wanted to convey and the idea the second person understood.

4. Error Rate

The amount of words, ideas, relationships and conclusions that were identified incorrectly.

5. Receiver Comprehension

A measure of whether the person who received the information correctly understood its true meaning.

6. User Effort

The attention, mental strain or fatigue, training and ease of use required to use the technology.

7. Privacy Protection

Monitoring whether thoughts or information the user has not authorised are accidentally disclosed.

A technology cannot be said to have accelerated communication between people merely because its network connection is fast.

Likewise, the fact that artificial intelligence generates a sentence very quickly is not sufficient evidence.

The complete human-to-human communication process must be measured.

7. The Proposed Experimental Research Plan

This research can be carried out in three main stages.

Stage One: Identifying Human Intentions

Initially, it must be investigated whether a few specific ideas that people deliberately wish to convey can be identified from brain signals.

Already published EEG datasets, or new studies conducted with proper consent and permission, can be used.

For example, participants may attempt to convey one of the following ideas.

• I agree.

• I do not agree.

• I need help.

• I need further explanation.

Machine-learning models can be used to examine whether differences appear in the brain signals associated with these ideas.

The model's performance must then be evaluated using data not used in training, different test sessions and independent participants.

The important question at this stage:

Is it possible to identify reliably, from brain signals, the idea a person wishes to convey?

Stage Two: Meaning Transfer with the Help of Artificial Intelligence

An attempt can then be made to convert the information identified in stage one into complete ideas with the help of artificial intelligence.

Three aspects must be compared here.

First, the idea the sender actually wanted to convey.

Second, the idea the artificial intelligence reconstructed.

Third, the meaning understood by the person who received that information.

How accurately these three correspond must be examined through mathematical and statistical methods.

Most importantly, when an idea is not clear to the artificial intelligence, it must express its uncertainty rather than generate incorrect information.

The performance of this method must be compared with ordinary typing, speech and other communication methods.

Stage Three: Communication Between People

In the third stage, information-transfer experiments can be conducted between two people participating with full consent.

The first person deliberately attempts to convey information through a brain signal.

Artificial intelligence analyses that signal, identifies the relevant information and sends it securely.

The second person receives and understands that information.

The speed, accuracy and reliability of the complete communication process are then measured.

Delivering information directly to the recipient's brain would be a much more advanced and separate research stage.

Such attempts require neural expertise, regulatory approvals, medical safety and strict ethical controls.

Priority should therefore be given, in early-stage research, to proving whether artificial intelligence can improve the performance of communication between people.

8. Scientific Challenges and Current Limitations

There are several important challenges in this research.

First, not all human brains represent words and ideas in the same neural patterns.

Each person may therefore need individual training or model adjustment.

Second, signals recorded in non-surgical methods such as EEG may contain high levels of noise and interference.

Compared with electrodes implanted in the brain, the ability to detect certain kinds of neural activity in fine detail may also be lower.

Third, attempting to speak, imagining speech in the mind, intending to express an idea and thinking privately to oneself are not the same kind of neural activity.

These must be studied separately.

Fourth, large language models can sometimes generate sentences that are untrue but appear plausible, based on the information available to them.

In human thought-exchange technology, such errors can become a very large problem.

Telling someone that a person thought something they did not think can create serious misunderstandings.

Fifth, it is very important that the results of scientific studies can be reproduced by other researchers.

A study published in the Journal of Neural Engineering in 2026 showed that there are many problems in reproducing the results of studies that identify imagined speech.

Among the speech-imagery datasets examined in that study, only 36% of participants achieved classification accuracy that crossed the appropriate statistical significance threshold [10].

This demonstrates how important independent verification, transparent research methods and reliable testing are in this field.

Finally, the speed of information transfer is not the only challenge.

Human attention, memory, comprehension and thought processes also play important roles.

Removing the need to speak does not, by itself, make the time needed to form or understand a complex idea disappear entirely.

Based on current scientific evidence, it cannot be said that every kind of human thought can be transferred directly at gigabit speed.

Treating digital data-transfer speed and the transfer speed of meaningful human thoughts as the same thing is a scientifically incorrect approach.

9. Privacy, Information Security and Human Rights

Technologies that interpret the neural signals of the human brain also create very important ethical responsibilities.

In ordinary computer systems, the information processed is information that people enter themselves.

But neural interfaces may also record signals associated with a person's private mental activity that they do not wish to share with others.

The most important basic principle of our research must therefore be the following.

Only the information that a person, with their full consent, deliberately wishes to share should be identified and sent to others.

Several safety arrangements are necessary to put this principle into practice.

The user must have explicit control to initiate communication.

Wherever possible, neural data should be processed on the user's own device.

Information must be transferred in encrypted form.

Strong security systems must be created to prevent unauthorised persons from accessing brain signals.

Storing unnecessary neural data for long periods must be avoided.

The user must have the facility to stop the transfer of information immediately at any time.

In addition, it must be made clear that the ideas generated by artificial intelligence are not one-hundred-percent certain records of the thoughts in a person's mind.

They are estimates generated on the basis of neural signals.

In November 2025, UNESCO adopted an international recommendation on the ethical use of neurotechnologies [11].

That recommendation gives importance to mental privacy, human dignity and the responsible use of neural technologies.

In our research too, privacy and ethical safeguards must proceed together with technological development.

10. Future Applications of This Technology

If this research succeeds, its early benefits may reach people who face severe difficulties with speech and physical movement.

In particular, brain-computer interfaces may create new opportunities for those who cannot communicate because of motor paralysis and certain neurological conditions.

Such technologies may also be used in some special situations where communicating through speech or body movements is difficult.

They could also improve computing accessibility for people with disabilities.

In the longer term, this research may also help to make communication between people and artificial intelligence systems more efficient.

But the real benefits of every such application must be proven scientifically, one by one.

It cannot be assumed in advance that direct neural communication will be better than ordinary human conversation in every situation.

Human language is not merely a slow tool for transferring information.

It also performs many important functions: thinking, reasoning, emotional expression, clarifying ideas and building human relationships.

New technology should therefore enhance people's communication capacity, not make them lose these important qualities of human language.

11. The Scientific Contribution of This Research and the Path Ahead

This research is not proposed on the assumption that a technology that transfers all human thoughts directly has already been created.

Instead, it is proposed as a scientific research framework that brings together the identification of brain signals, understanding the meaning of ideas with the help of artificial intelligence, measuring the performance of information transfer and evaluating people's genuine understanding.

Its important question:

Is not whether the full data-transfer capacity of an electronic network can be created inside the human brain.

The important question is whether the time between the moment one person forms an idea in their mind and the moment another person correctly understands that idea can be reduced.

For this, it must be investigated whether meaning-based information representations created with the help of artificial intelligence provide additional performance compared with neural technologies that identify ordinary speech.

It is not enough, to say a technology is successful, that artificial intelligence generates elegant sentences.

It must reduce the time information transfer takes.

It must protect the true meaning of the idea being sent.

It must control errors to an acceptable level.

It must work reliably with different people and in practical settings.

These conditions must be proven through scientific experiments.

Research must also accept that some approaches may fail to perform better than ordinary communication.

The true value of scientific research lies not only in proving an expected result.

It lies in discovering, on the basis of reliable evidence, whether the hypothesis is right or wrong.

12. Conclusion

Over the past few decades, computer data-transfer technologies have advanced enormously.

Today, computers located in different parts of the world exchange very large amounts of information at high speed.

But communication between people still largely depends on speech, writing and other physical forms of expression.

Recent neural research has produced notable advances in reconstructing speech from brain signals, identifying speech imagined in the mind, and transferring defined information between brains.

Nevertheless, a technology that transfers every kind of human thought directly, accurately and at high speed has not yet been created.

It remains a great scientific challenge to be solved.

In the future, the integration of artificial intelligence and neural technologies may create new ways to reduce the time people need to express their thoughts.

Practical research for this must begin from the neural signals people deliberately try to convey.

Models that identify them accurately must then be built.

Next, it must be assessed how accurately the true meaning of the information is protected.

Finally, it must be proven scientifically whether this technology genuinely improves the speed and performance of communication between people.

Only then can work move forward towards more advanced technologies such as complete direct brain-to-brain communication.

The Ultimate Goal of My Research

The ultimate goal of my research is to reduce as far as possible the gap between the speed at which a person forms a thought in their mind and the speed at which another person accurately understands the true meaning of that thought.

For this, my proposed approach is to study artificial intelligence, brain-computer interfaces, neuroscience, information theory and secure high-speed communication technologies together.

Our goal is not merely to send more data at high speed.

Our goal is to build the scientific foundation for a new communication method that can transfer people's thoughts and ideas safely, accurately and in the least possible time, without changing their true meaning.

References

1. Coupé, C., Oh, Y. M., Dediu, D., & Pellegrino, F. (2019). Different languages, similar encoding efficiency: Comparable information rates across the human communicative niche. Science Advances, 5(9). https://doi.org/10.1126/sciadv.aaw2594
2. Zheng, J., & Meister, M. (2025). The unbearable slowness of being: Why do we live at 10 bits/s? Neuron, 113(2), 192–204. Published online: December 2024. https://doi.org/10.1016/j.neuron.2024.11.008
3. Willett, F. R., et al. (2023). A high-performance speech neuroprosthesis. Nature, 620, 1031–1036. https://doi.org/10.1038/s41586-023-06377-x
4. Littlejohn, K. T., et al. (2025). A streaming brain-to-voice neuroprosthesis to restore naturalistic communication. Nature Neuroscience, 28, 902–912. https://doi.org/10.1038/s41593-025-01905-6
5. Wairagkar, M., et al. (2025). An instantaneous voice-synthesis neuroprosthesis. Nature, 644, 145–152. https://doi.org/10.1038/s41586-025-09127-3
6. Card, N. S., et al. (2026). Long-term independent use of an intracortical brain–computer interface for speech and cursor control. Nature Medicine, 32, 2504–2510. https://doi.org/10.1038/s41591-026-04414-6
7. Jiang, L., et al. (2019). BrainNet: A Multi-Person Brain-to-Brain Interface for Direct Collaboration Between Brains. Scientific Reports, 9, Article 6115. https://doi.org/10.1038/s41598-019-41895-7
8. Kunz, E. M., et al. (2025). Inner speech in motor cortex and implications for speech neuroprostheses. Cell, 188(17), 4658–4673.e17. https://doi.org/10.1016/j.cell.2025.06.015
9. Zhao, Z., et al. (2026). A neural architecture for imagined and overt speech motor dynamics. Nature Neuroscience. https://doi.org/10.1038/s41593-026-02456-0
10. Tates, A., Matran-Fernandez, A., Halder, S., & Daly, I. (2026). Decoding speech imagery or just noise?: A symptom of the replicability crisis. Journal of Neural Engineering, 23(5). https://doi.org/10.1088/1741-2552/aea294
11. UNESCO. (2025). Recommendation on the Ethics of Neurotechnology. https://www.unesco.org/en/node/86248

By Keshavan Thiagaraja

Powered By K7