OpenAI’s Chief Scientist Suggests Perplexity Could Help Limit AI Development

TechCrunch

To establish a speed limit on AI development, you need to agree on how to measure speed.

OpenAI’s chief scientist, Jakub Pachocki, has published a new blog post exploring potential limits on AI development, focusing on a measurement known as perplexity.

The question of slowing AI development has become a major topic of discussion after a former researcher from OpenAI and Anthropic claimed that the companies were “gambling with our lives.” Anthropic’s initial public offering filing also suggested that the technology could pose “catastrophic or existential risks to society.”

Anthropic CEO Dario Amodei subsequently released a plan to “pace the frontier,” which appeared to receive support from OpenAI CEO Sam Altman and SpaceX CEO Elon Musk. Amodei’s proposal included a call for companies to agree on “common safety standards as well as limits on the rate of unchecked AI progress.”

However, setting a speed limit for AI development requires a shared understanding of how to measure speed. Pachocki’s post—which Altman described as “the most important thing OpenAI has published in a while”—suggests that perplexity could serve as that measurement. This refers to the technical concept and not the company that shares the same name.

Pachocki argues that benchmarks focused on AI’s “downstream capabilities” quickly become outdated as the technology advances. Perplexity, which “measures how well a model predicts the next token on average; lower perplexity means better predictions,” has remained “informative.”

The post includes a chart showing how the perplexity of OpenAI’s models has declined since GPT 3.5. Pachocki wrote that GPT-4, at approximately 0.91, was proficient at using chain-of-thought reasoning productively, which motivated the targeted elicitation of this capability and the creation of modern reasoning models. He added that as losses dropped below 0.8, transformative capabilities emerged in fields such as cybersecurity and mathematics.

👉👉👉  Школьные олимпиады и тесты онлайн: как выбрать задания для ребенка

In Pachocki’s view, perplexity could provide the basis for “a shared benchmark for general intelligence,” which AI companies could use to establish “common limits.” With perplexity as a standard measurement for AI progress, companies would commit not to exceed a certain threshold “until our ability to understand and monitor AI behavior achieves a certain concrete milestone.”

Source: TechCrunch

To establish a speed limit on AI development, you need to agree on how to measure speed.

— TechCrunch

Deja una respuesta

Tu dirección de correo electrónico no será publicada. Los campos obligatorios están marcados con *