From: mk_thisisit
According to physicist Roger Penrose, the common understanding and naming of “artificial intelligence” (AI) are fundamentally flawed. He asserts that true intelligence necessitates consciousness, a quality he believes current computational machines cannot achieve [00:00:04].
The Nature of Intelligence and AI
Penrose’s core thesis is that intelligence requires consciousness [00:00:07], and current computers, being purely computational machines, cannot gain consciousness [00:00:24]. He maintains that consciousness itself is not computational [00:00:30].
“The name is wrong; it is not artificial intelligence. Intelligence requires consciousness.” [00:00:01]
He suggests that a more accurate term for what we currently call AI might be “artificial cleverness” [00:14:09]. This distinction highlights that while AI can perform tasks cleverly and powerfully [00:09:02], it does not possess understanding or awareness of its actions [00:00:35]. Computers have become so powerful that people often “lose sight of what they are doing” [00:00:42].
The Turing Test Reconsidered
Alan Turing’s idea of a “Turing test” to determine if a being is conscious through conversation [00:09:51], while influential, is seen by Penrose as missing a crucial element [00:10:02].
The Limits of Computation
Penrose argues that computers are based on computational mathematics, which is a very limited part of mathematics [00:01:12]. He claims that artificial intelligence “simply perform calculations” [00:02:23] and are essentially “simple statistical machines” [00:11:01].
“Computers are a special type of mathematical structure based on computational mathematics, which is a very limited part of mathematics.” [00:01:12]
Gödel’s Theorem and Uncomputability
Penrose’s views on computation and consciousness were shaped by his studies in mathematical logic at Cambridge, where he learned about Turing machines, computability, and Gödel’s theorem [00:03:12]. Gödel’s theorem indicates that there are mathematical truths whose understanding “goes beyond their use” [00:03:31].
Penrose explains Gödel’s theorem as the ability to construct a statement whose meaning a person can understand, but which cannot be proven using a given set of computational rules [00:06:02]. The core of this is the inability of a computational system (like AI) to know if its rules are true [00:07:13].
“Because [AI] doesn’t know that they are true. This is the whole point of Theorem G, in which it is about how to go beyond the rules. This is done by understanding why they are true.” [00:07:13]
This understanding requires awareness, a quality that allows humans to transcend computational rules [00:08:14]. Mathematical logic reveals the existence of “uncomputable things” and “uncomputable structures” [00:05:02]. Penrose believes that part of mathematics is not subject to calculation [00:16:21].
Consciousness and Physics
As a physicist, Penrose believes that whatever creates consciousness belongs to the physical world, not magic [00:11:33]. However, he posits that the physics related to conscious thinking cannot be calculated [00:12:16]. In contrast, much of the physics we currently understand, such as general relativity and the dynamics of black holes, can be calculated [00:12:29].
Penrose speculates that true consciousness might be connected to a “special type of new physics” [00:12:04] which is “in principle uncalculated” [00:12:57]. This means current AI, being based on computers, “excludes consciousness” [00:13:44].
Quantum Reality and Backward Causality
Penrose suggests that consciousness might be connected to physics we do not yet fully understand, potentially stemming from or going beyond the quantum world [00:18:51]. While the quantum world is still considered computable, the “uncomputable” element concerns a “fundamental unknown” in the quantum world [00:19:00].
He argues that quantum theory is fundamentally incomplete, assuming that quantum superpositions hold at all levels [00:19:10]. He believes consciousness is connected with the “collapse of the wave function” [00:23:57], a mysterious phenomenon connected with “backward causality” [00:24:04].
Penrose distinguishes between:
- Classical reality: Determinable and understandable through our senses [00:24:49].
- Quantum reality: Can only be “confirmed,” not determined [00:26:13]. It behaves “very peculiarly” and operates on the principle of “backward causality” [00:27:17].
He references experiments related to the EPR paradox (Einstein, Podolsky, Rosen) [00:27:36], noting that while signals cannot be sent faster than light, one can conceptually “go back faster than light” in quantum reality [00:29:02]. Penrose concludes that until we understand how the quantum world truly works, we cannot transfer it to digital space [00:29:17].
Nobel Prize for AI Creators
Regarding the Nobel Prize in Physics being awarded to creators of AI, Penrose expresses skepticism, questioning if it is truly physics [00:20:14]. He considers such an award more indicative of “technological progress” rather than purely theoretical or fundamental physical studies [00:20:41].
Predictions and Risks
Penrose believes that current AI, despite its increasing power, will “not ever gain consciousness because they are machines” [00:22:51]. However, he acknowledges that AI will likely surpass human capabilities in many tasks [00:23:01].
The greatest risk in AI development is the widespread misconception that machines are, or can become, conscious [00:23:21]. If machines were to gain consciousness, the risk would be even greater [00:23:27]. Penrose maintains that consciousness is a physical process that we do not yet fully understand [00:23:49], hinting that its true nature lies much deeper than current computational models can grasp [00:18:36].