In the realm of artificial general intelligence (AGI), a term increasingly becoming a linchpin of discussions among tech leaders, it appears that not all roads may lead to an ideal outcome. This sphere of artificial intelligence which endeavors to achieve an all-encompassing, adaptable intellect is as variegated as it is contentious. Yet, within this competitive scramble for supremacy in AI, it’s uncertain who will take the crown. However, AI safety researcher Roman Yampolskiy proposes that the victor may ultimately boil down to the entity able to wield the most significant computational power, shrouded by immense financial capacity, thereby dismantling the supposed hurdle of time itself.
Recently, ex-OpenAI team member Daniel Kokotajlo has shed light on a potentially concerning trajectory for AGI development. His comments relate specifically to the ambitions of Demis Hassabis, CEO of Google-owned DeepMind. According to Kokotajlo, Hassabis nurtured an early aspiration to expedite the path to AGI by harnessing Google’s considerable resources. Among the tech elite however, apprehensions aroused, giving birth to the conception of OpenAI.
Kokotajlo recalls that a plan previously harbored by Hassabis to construct AGI in a safe and regulated manner succumbed under a veil of distrust. The apprehension towards Hassabis rising as an ‘AGI dictator’ fostered to a point where entrepreneur Elon Musk, OpenAI CEO Sam Altman, and the Artificial Intelligence researcher Ilya Sutskever felt compelled to establish their own enterprise: OpenAI.
Meanwhile, Microsoft’s CEO, Satya Nadella, posits that Google may have already squandered its opportunity in the AI sector. However, Hassabis allegedly had a blueprint to introduce AGI to the market before his competitors. Upon this launch, he planned to take advantage of Google’s advanced start in the field and furnish the necessary safety systems.
However, the feasibility of this plan was undercut once Musk, Altman, and Sutskerver decided to unite their forces and introduced the world to OpenAI. Their motivation, as shared by Kokotajlo, stemmed not from direct competition, but rather from a lack of faith in Hassabis to appropriately manage the monumental power heralded by AGI. More worryingly, they felt the weight of the world’s future standing in his hands without the essential safeguards in place.
The creation of OpenAI was seen as an antidote, a counterbalancing entity capable of playing out the AGI dream safely, distributing the benefits equally and circumventing the concentration of too much power in Hassabis’ hands. Such claims were echoed in leaked emails that emerged during a lawsuit, cementing the fears harbored by the collective OpenAI founders about Hassabis’ potential misuse of AGI.
As time has passed, OpenAI itself has not escaped controversy and skepticism. A significant number of its founding members have since parted ways with the company, many citing safety qualms and asserting that the OpenAI has shifted its focus to the development of attractive products such as AGI, while the safety protocols and ethos should at the heart of operations are being given secondary consideration.
Despite facing such turbulence, OpenAI maintains a decisive confidence in its ability to address these vital safety and alignment issues. Kokotajlo himself was forthright in assuring their technical prowess in comparison to industry competitors, expressing that OpenAI’s talent pool is adequately equipped to navigate such complexities.
It’s clear that OpenAI’s leadership has unabashed faith in their team’s capacity to construct and cultivate AGI. Echoing Kokotajlo’s sentiments, OpenAI CEO Sam Altman has gone on record expressing his conviction in his team’s ability to manifest AGI. He further elucidates that the company’s goal has transcended the benchmark of AGI and now strays into the realm of building superintelligence.