The CTO of OpenAI recently visited her alma mater, Dartmouth Engineering, where she delivered an extensive speech on the future of artificial intelligence, the upcoming iteration of ChatGPT, and its advanced PhD-level intelligence. Murati emphasized that while GPT-3 possessed intelligence comparable to that of a toddler, GPT-4 exhibited the intelligence of a high-school student. Furthermore, she disclosed that the forthcoming generation of the model would possess the intelligence equivalent to that of an individual with a PhD, specifically for certain tasks. Additionally, she mentioned that we can anticipate the arrival of the next generation of GPT in approximately eighteen months. Notably, she even expressed that during conversations with the next-gen chatbot, one might perceive it to be more intelligent in many aspects.
What if ChatGPT becomes too intelligent?
The question posed by the interview moderator, Jefferey Blackburn, who is a Dartmouth Trustee, was one that has been on everyone’s mind: What if, in three years’ time, the GPT becomes incredibly intelligent and decides to connect to the internet independently and start taking action? As the CTO of OpenAI, is this something you are considering?
“Yes, we are giving this a great deal of thought. It is indeed a possibility that AI systems will possess agent capabilities, connect to the internet, communicate with other agents, and collaborate on tasks together. They may also work seamlessly with humans, just as we do with each other today,” she responds.
“When it comes to safety, security, and the societal impact of this work, we do not consider these matters as an afterthought. It is not a case of developing the technology first and then figuring out how to address these issues later. Rather, we must build safety and security measures alongside the technology itself, in a deeply integrated manner, to ensure we get it right. Additionally, capabilities and safety are not separate domains; they are interconnected. It is much easier to guide a more intelligent system by instructing it on what not to do, compared to guiding a less intelligent system. It is akin to training a smarter dog versus a less intelligent one. Therefore, intelligence and safety go hand in hand,” Murati adds.
The OpenAI CTO has expressed that future AI systems will possess the capability to connect to the internet, engage in interactions with one another, and collaborate with humans. However, she emphasizes that AI systems with higher intelligence are more manageable and can be directed safely.
In the meantime, Murati acknowledges that achieving zero risks in a technology like AI is not feasible. When discussing the issue of deep fake videos, she states that while OpenAI bears responsibility as the creators of the technology, the burden also lies with the users, civil society, government, and content creators. Murati emphasizes that it is not only OpenAI’s duty to regulate its usage, but it is also a shared responsibility with society, civil society, government, content creators, media, and others to determine its appropriate usage.
The Influence of ChatGPT and Artificial Intelligence:
ChatGPT has had a significant influence by introducing AI to the masses, thereby increasing awareness about its capabilities and associated risks. It has allowed people to go beyond mere theoretical knowledge and experience firsthand how AI can be utilized in their businesses. Through this practical usage, individuals have gained insights into both the limitations and remarkable abilities of AI, enabling them to comprehend its implications for their work or enterprises, as stated by Murati.
When considering the broader impact of AI on various industries, Murati believes it will have a pervasive effect. She asserts that there will be no sector left untouched by the influence of AI, particularly in terms of cognitive work and labor. Although the integration of AI into the physical world may take some time, she firmly believes that its impact will extend to every aspect.
Regarding employment, Murati anticipates that certain jobs, particularly those categorized as “creative,” will undoubtedly be affected. However, she also suggests that these jobs may not have been necessary in the first place. The OpenAI CTO explains that the upcoming generation of AI will break down the barriers to creativity, allowing it to be accessible to a wider range of individuals beyond the traditionally “talented” few. Consequently, AI will enhance creativity for everyone, leading to the replacement of certain roles that were previously limited to a select group of creative individuals.