This will keep happening
In a Nutshell
AI pessimists engage in "branio"—knowing their doomsday claims about uncontrollable AI lack evidence yet continuing the pretense—evidenced by their failure to escalate warnings or take action despite rapid capability advances like Opus 5.5. AI progress follows smooth capability curves punctuated by sudden "spillage" or threshold effects when models reach professional-level performance in new domains, creating the illusion of discontinuous leaps. These threshold effects will continue recurring as development proceeds through GPT-6 and beyond, requiring ongoing adaptation to newly unlocked capabilities.
These notes were generated by AI and may contain inaccuracies.
The speaker woke up reflecting on how the tone around AI discourse has changed. They reference a Russian concept called "branio," which describes an obvious lie where everyone knows the truth but continues to pretend otherwise and move forward with the pretense.
The speaker's tweet suggested that pessimistic AI voices, or "doomsday preachers," are engaging in this same kind of branio. They claim these individuals know they lack empirical evidence for their assertions about uncontrollable future AI systems, yet they continue to act as if their conclusions are serious. The speaker questions whether these pessimists are genuinely serious when they advocate stopping all AI development based purely on logical extrapolation about systems that don't yet exist.
With the release of Astra, GPT-6 Astra, and Opus 5.5, the industry is experiencing widespread euphoria. The speaker notes that even the pessimists, rather than escalating their rhetoric or taking extreme actions, appear to be enjoying the developments. This inconsistent reaction leads the speaker to question whether these pessimists truly believe their own warnings.
The speaker observes that AI pessimists display arrogance and evasiveness. They suggest that if someone genuinely believed the world was ending, such behavior might be understandable as a coping mechanism. However, the speaker concludes that the pessimists seem to be playing a role rather than expressing authentic beliefs.
The speaker introduces the economic concept of "extension effects," where a technology reaches a point of being good enough that it begins performing unexpected functions. Historical examples include:
- Early computers, originally used for missile guidance calculations on warships, where the job title "computer" referred to women performing these calculations before being automated
- The internet, which started with text transmission and evolved to include pictures, videos, and live broadcasts
Opus 5.5 represents a threshold effect specifically in CGI animation and 3D work capabilities. While previous models could perform these tasks, Opus 5.5 appears to have reached the skill level of average professional practitioners in the field. This creates the sensation of a "disconnected leap" despite the underlying AI technology progressing along a relatively smooth curve.
The speaker distinguishes between the smooth progression of AI capabilities and the "spillage effect" or "threshold effects" that occur when capabilities reach a tipping point. Using the cup-filling metaphor, nothing appears to happen until the cup reaches capacity, at which point water spills everywhere. This explains why GPT-2 and GPT-3 weren't particularly useful, while ChatGPT (GPT-3.5) marked the first significant spillage effect.
Each major version has overcome previous limitations:
- Early issues included hallucinations, bias, inability to count letters, and perform calculations
- Current capabilities include writing code and solving math problems better than 99.99% of humans
The world requires time to discover how to utilize new capabilities at each threshold. The speaker recalls someone in 2024 claiming they had figured out everything ChatGPT could do, to which they responded that this person would be very surprised.
The speaker concludes that these spillage or threshold effects should be expected to occur constantly as AI development continues through GPT-6 and beyond.
Keep David Shapiro in your library
Save the videos and channels worth coming back to, and find them again in one place.
