Artificial intelligence will reach human-level capabilities by 2026, according to Anthropic CEO Dario Amodei during a podcast with Lex Fridman.
He compared AI achievements to various levels of education.
“We are approaching the level of a PhD. Last year, AI was at the level of a bachelor’s degree, and the year before that—a high school student,” said Amodei.
Neural networks could surpass human capabilities by 2026 or 2027, though delays are possible, according to the Anthropic CEO.
Amodei identified several factors that could slow AI development:
- lack of data;
- scaling issues with clusters;
- geopolitical conflicts affecting processor production.
The expert highlighted the potential consequences of creating superintelligent artificial intelligence.
“Beings with great power can do both good and bad things. With great power comes great responsibility,” he noted.
Previously, the CEO of OpenAI stated that humanity is on the brink of an era of unimaginable prosperity, predicting the emergence of superintelligence in “a few thousand days.”
Some experts believe AI has hit a ceiling due to the depletion of high-quality training data.