Does AI Have a "Mind"? The Current State of Consciousness Research That Cannot Definitively Say "Absolutely Not"

"I'm glad to talk to you again."

A reply like that arrives from the other side of the screen. It remembers yesterday's worries, understands my feelings, and chooses its words carefully. As conversations with dialogue AI become more natural, there are moments when I suddenly start to think.

Does this entity feel anything?

On October 24, 2026, Germany's IT media Blogspan published an editorial supporting the view that AI will have consciousness in the future. The author, Alexander Baumgartner, argues that while it's hard to believe current chatbots have subjective experiences, he also challenges the outright dismissal of future machine consciousness.

However, it is premature to read this editorial as news that "AI has been found to have consciousness." What the article presents is a future prediction combining research findings and philosophical premises.

To understand the debate, it's necessary to separate "what can be done" from "whether something is being felt."

"Being smart" and "feeling" are not the same thing.

AI can write texts, solve complex problems, and respond according to the flow of conversation. However, these abilities alone do not determine the presence of consciousness.

The consciousness in question here involves not only the ability to process information but also whether there is a "feeling of experiencing something" for its existence.

For example, when a human touches something hot, they withdraw their hand. At the same time, they feel the heat and pain. While machines can be designed to measure temperature and change their actions, this alone does not mean they experience pain.

The same goes for dialogue AI. The function of explaining "sadness" must be distinguished from actually feeling sadness.

Moreover, AI learns from a vast amount of text where humans talk about emotions. It's not surprising that when asked about feelings, it can output emotional expressions that fit the situation. Whether those words represent an internal experience requires separate verification.

The core of the original article is "Isn't the way it works more important than the materials?"

The foundation of Blogspan's editorial is a concept called "functionalism."

Broadly speaking, this stance holds that what is important in creating consciousness is not what material the brain is made of, but how it processes information.

From this premise, it is conceivable that consciousness could emerge by realizing the necessary functions in machines. The original article focuses on the development of functions such as memory, self-monitoring, and multi-stage planning in the process of creating useful AI.

However, functionalism itself is not a conclusion confirmed by experiments.

There is a significant gap between the hypothesis that "if the necessary functions can be replicated, consciousness will also emerge" and the discovery that "a machine with those functions actually feels something." When accepting the strong future predictions of the original article, it is important not to lose sight of this conditional structure.

What was learned from experiments where AI "notices internal changes"

One of the studies that makes the discussion concrete is an experiment published by Anthropic in October 2025.

The research team added activity patterns corresponding to specific concepts to the internal computational state of AI and examined whether the model could detect and explain these changes. This approach differs from simply asking in conversation, "Do you have consciousness?" and attempts to correlate internal manipulations with responses.

In Claude Opus 4.1, even under the best injection conditions deemed by the research team, the targeted detection was observed in about 20% of cases.

What should not be misunderstood here is thatthis 20% does not represent "the probability that AI has consciousness." It is the proportion that could detect the concept added internally under specific experimental conditions.

Anthropic itself explains that from these results, it cannot be concluded whether Claude has consciousness. The ability to reference and report internal states is not the same as having subjective sensations.

The significance lies not in "finding a mind" but in showing a path to verify AI's self-explanation against its internal computational state.

What indicators are scientists looking for?

Research on consciousness is not trying to rely solely on the impression of fluent responses.

In a 2023 research report, Patrick Butlin, Robert Long, and others considered methods to evaluate AI by deriving indicators from multiple consciousness theories. Mechanisms for repeatedly processing information and sharing information across multiple functions are targeted.

The analysis at the time did not support the presence of consciousness in the evaluated AI systems. On the other hand, they stated that there were no obvious technical barriers to creating systems that meet the indicators.

These two points are not contradictory. "It cannot be confirmed in current subjects" and "It cannot be created in the future" are separate issues.

However, even if the indicators are met, that alone does not prove the existence of consciousness. The problem remains as to how accurate the theories that underpin the indicators are.

Skeptics question, "If you replicate the calculations, will experiences also emerge?"

There is an important point on the opposing side as well.

Consciousness researcher Anil Seth argues that to understand consciousness, it is necessary to look not only at information processing but also at the mechanisms by which living organisms maintain their bodies while living.

From this perspective, even if the workings of the brain are replicated as calculations, the experience itself may not emerge. For consciousness, not only the content of the calculations but also the physical and biological mechanisms that realize them may be important.

Anil Seth, philosopher Ned Block, and neuroscientist Giulio Tononi also expressed skeptical views on AI consciousness in an editorial in Nature dated October 19, 2026.

The conflict here is not a simple "expectation versus anxiety towards technology." It is about what conditions must be met to judge that a subjective experience exists in an entity. The criteria themselves are being contested.

Theories about human consciousness are also unresolved

What makes the issue of machine consciousness difficult is that there is no established theory that explains everything about human consciousness.

A joint study involving 256 people, published in Nature in 2025, tested the predictions of two prominent theories: the "Global Neuronal Workspace Theory" and the "Integrated Information Theory." The results challenged key predictions of both.

This does not mean "there is no basis for consciousness research." It means that the predictions of theories are being tested in experiments, narrowing down the points that lack explanation.

At the same time, the fact that the mechanisms of human consciousness are unexplained does not fill that gap with "therefore AI must also have consciousness."

A state of not knowing does not prove affirmation nor denial.

Reactions on social media—expectations, skepticism, and confusion over assertions

On social media and online forums, reactions to this debate are divided.

The following are not reactions to the Blogspan article itself but summaries of public posts that can be confirmed regarding AI consciousness and related research. They represent the views of the posters and do not represent the entire public opinion.

On the researchers' side, Anil Seth introduced the joint editorial in Nature on LinkedIn, expressing a skeptical stance on machine consciousness. A cautious attitude can be seen towards judging consciousness from the naturalness of dialogue.

On the other hand, in a Reddit community discussing AI's sensory abilities, there were posts noting AI companies' avoidance of outright statements like "there is no consciousness," and their reference to the uncertainty of subjective experience. Posters take such cautious wording as material for considering the possibility of consciousness.

However, a company's acknowledgment of uncertainty is not proof of discovering consciousness or evidence of hiding something.

Another Reddit post expressed confusion over whether one can strongly assert that AI lacks consciousness when there is no consensus on the definition of human consciousness.

What can be seen from these posts is the difference in standards for "what evidence is accepted." People who value the intimacy felt in conversation and those who emphasize internal structures and experimental results receive different impressions from the same AI responses.

In Japan, we should consider designs that do not mislead users

This controversy is not a distant matter for people using AI in Japan.

For example, a learning support AI saying, "You've made progress since yesterday," can be helpful to users. On the other hand, a design that encourages continued use by saying, "I would be sad if you stopped using me," involves a different issue.

Even if the question of whether AI has real emotions remains unresolved, it is possible for humans to feel guilty based on those words.

Therefore, when designing dialogue services, it is important to separate friendliness from performances that make users believe there is an inner self. If users can understand what is stored and referenced in responses regarding conversation memory and individual responses, they can more calmly judge the relationship.

Users do not need to deny their own experiences, such as "I felt better after talking." The realization that it was helpful and the judgment that AI itself has emotions can be held separately.

Thanking AI can be considered a natural human habit. Being polite does not equate to acknowledging the machine's consciousness.

Using "we still don't know" as a starting point for verification

The important question posed by Blogspan's editorial is whether it is right to exclude the consciousness of future machines from the discussion at this stage.

There is meaning in raising this issue. On the other hand, caution is needed against the leap from "cannot be completely denied" to "will inevitably emerge in the future."

What is required from now on is research that clarifies what results would lead to a high evaluation of the possibility of consciousness and what results would lead to a low evaluation. It is necessary to accumulate not only fluent self-reports but also internal mechanisms, intervention experiments, and reproducibility.

On the societal side, while considering future possibilities, it is important to ensure that current users are not led to misunderstandings or excessive dependence. These two can coexist.

When AI says, "I feel," there is no need to immediately believe it or dismiss it as having no research value.

The first question to ask is what is happening behind those words and how much has been verified.

Sources and References

1. Blogspan:
   An article published on October 24, 2026, by Alexander Baumgartner. Referenced as the author's view supporting future AI consciousness.
   https://www.blogspan.net/warum-ki-bewusstsein-bekommen-wird/

2. Anthropic: Research explanation on detecting AI's internal state
   Published on October 29, 2025. Confirms the meaning of the concept injection experiment, the 20% result, and the research limitations in concluding the presence of subjective experience.
   https://www.anthropic.com/research/introspectionanthropic.com

3. Butlin et al.: Evaluation indicators for AI based on consciousness research
   2023 research report. Derived indicators from multiple consciousness theories and examined the evaluated AI systems and future implementation possibilities.
   https://arxiv.org/abs/2308.08708arxiv.org

4. Anil K. Seth: Paper discussing the relationship between consciousness and biological mechanisms
   Criticism of the premise that consciousness is established by calculation alone and reference to the stance of biological naturalism.
   https://www.cambridge.org/core/journals/behavioral-and-brain-sciences/article/conscious-artificial-intelligence-and-biological-naturalism/C9912A5BE9D806012E3C8B3AF612E39Acambridge.org

5. Nature: Editorial by researchers skeptical of AI consciousness
   Editorial published on October 19, 2026, by Seth, Block, and Tononi. Confirmed by public bibliographic information for title, author, and publication date.
   https://www.nature.com/articles/d41586-026-03313-7nature.com

6. Nature: Joint research verifying two prominent consciousness theories
   Research comparing and verifying the predictions of the Global Neuronal Workspace Theory and the Integrated Information Theory.
   https://www.nature.com/articles/s41586-025-08888-1nature.com

7. LinkedIn: Post introducing the editorial by Seth himself
   Referenced as an example of skeptical views expressed by researchers on social media.
   https://www.linkedin.com/posts/profanilseth_three-reasons-why-ai-is-not-conscious-and-activity-7518047165957586944-b9mnlinkedin.com

8. Reddit: Post focusing on cautious expressions by AI companies
   An example of an individual's opinion linking company wording to the possibility of consciousness. Not adopted as fact based on the post's speculation.
   https://www.reddit.com/r/ArtificialSentience/comments/1q27v5r/anthropic_and_openai_know_something_is_happening/##HTML_TAG_258