OpenAI Deciphers AGI

Genius Andrej Karpathy, who previously worked at Tesla and now works at OpenAI, said he is "building something like JARVIS at OpenAI". Undoubtedly, this guy who created Baby Llama and can easily write GPT-5 over the weekend may be ready to build autonomous agents, and if not AGI, OpenAI may have cracked it.

OpenAI co-founder Greg Brockman recently shared the capabilities of GPT-4 Voice, where the assistant can negotiate business deals with minimal supervision. Ironically, it identifies itself as Jarvis. Ammaar Reshi released a screen recording of a conversation with a chatbot, where the chatbot refers to itself as Jarvis and is ready to assist him in everything.

Undoubtedly, OpenAI has been committed to building AGI. Interestingly, the company has taken another step towards this goal and changed its "core values" on its website to increase focus on AGI, which was not explicitly mentioned on the previous page.

Earlier, OpenAI's recruitment page listed six core values of its employees, such as "bold, thoughtful, impact-driven, collaborative," etc. But now, the first one is AGI, followed by "intense and ambitious, scalable, making things people love, and team spirit".

Those who are confused about the company changing its "core values" may not realize that Altman and other co-founders have repeatedly stated that OpenAI is all about AGI.

OpenAI likely to announce "Jarvis" at DevDay conference

Some speculate that at the upcoming OpenAI DevDay conference, the company may announce their first fully autonomous agent. It is referred to as a "swarm of bees clicking on the internet" that fundamentally changes the field. This undoubtedly points to OpenAI's recent acquisition of Global Illumination, a company that trains AI agents through gamified simulations.

Although the conference focuses on developers and may include many developer-oriented announcements, the release of Jarvis seems to be already in progress.

From the rumors of autonomous agents, all recent developments indicate that OpenAI may announce something similar to Jarvis at the DevDay conference. In an interview, Altman defined AGI as "something like a middleman that you can hire as a colleague".

Autonomous agents are the next step based on the LLM chatbot. Almost all companies and researchers are aware of this. Microsoft recently launched a framework called AutoGEN, which allows building LLM applications using multiple agents that can interact with each other.

But as the AutoGEN GitHub repository states, "AutoGen agents can be customized, conversational, and seamlessly allow human involvement," which means these models still require human input, raising questions about the autonomy of the models.

Similarly, Google DeepMind recently published a paper titled "How far are large language models from intelligent agents with theory of mind?". Other similar papers include Meta's Shepherd: "Comments on language model generation" that discuss the same idea of enhancing autonomous AI agents and enabling them to complete tasks independently. Other related papers include SELF: a language-driven self-evolving large language model and SelfEvolve: a code evolution framework through large language models.

However, these researches have not been implemented in reality yet. If we mention OpenAI, they may have just cracked AGI like Jarvis.

All roads lead to AGI

Altman has always been prepared when it comes to the issue of unemployment. That's why he invested in Worldcoin and hopes to provide basic income for people. In that world, AI will do all our work, and we will sit at home focusing on building better things. Perhaps, the next one is Jarvis.

On the other hand, some have expressed concerns about the threat of autonomous killer robots, mentioning Ukraine's recent AI-based drones that automatically target and attack without human control, calling them "killer robots". As these AI-based drones may outperform human-operated drones, chaos is expected to arise in addition to job losses.

In another interesting note, have you ever received an email from a Nigerian prince trapped abroad asking you to send money? If you think these phishing emails are a problem, AI will provide a significant upgrade to them. Now, it's not ordinary people sending these emails, but an AI agent impersonating a Nigerian prince.