Silicon Valley Accelerates Development of Autonomous AI Agents

A new generation of AI assistants, equipped with advanced technology similar to ChatGPT and its competitors, is emerging, intensifying the competition in the field. This comes approximately ten years after the introduction of virtual assistants like Siri and Alexa. These new AI helpers possess enhanced autonomy, taking the capabilities of virtual assistants to a whole new level.

Experimental systems running on GPT-4 or similar models are attracting billions of dollars as Silicon Valley races to capitalize on advances in artificial intelligence. The new assistants — often called “agents” or “co-pilots” — promise to perform more complex personal and work tasks at the behest of a human without close supervision.

“High level, we want this to be something like your personal AI friend,” said developer Div Garg, whose company MultiOn is testing the AI agent.

“It could evolve into Jarvis, where we want this to connect to a lot of your services,” he added, referring to Tony Stark’s invaluable AI in the Iron Man movies. “If you want to do something, go talk to your AI and it will do your thing.”

The industry is still far from imitating the dazzling digital assistants of science fiction; Garg’s agent browses the web to order a burger from, say, DoorDash, while others can create investment strategies, email people selling refrigerators on Craigslist, or summarize work meetings for latecomers.

“A lot of what’s easy for humans is still incredibly difficult for computers,” said Kanjun Qiu, CEO of Generally Intelligent, an OpenAI competitor that creates artificial intelligence for agents.

“Say your boss needs you to book a meeting with a group of important clients. That involves complex reasoning skills for the AI—it needs to get everyone’s preferences, resolve conflicts, and at the same time maintain the careful touch needed when working with clients.”

According to interviews with Reuters, the early efforts are just a taste of the sophistication that could come in the coming years from increasingly sophisticated and autonomous actors as the industry moves toward artificial general intelligence (AGI) that can equal or surpass humans in countless cognitive tasks. two dozen entrepreneurs, investors and AI experts.

The new technology has sparked an onslaught of assistants running on so-called base models, including GPT-4, which has led to individual developers, major players such as Microsoft and Google parent Alphabet, and a host of startups.

Inflection AI, one startup, raised $1.3 billion at the end of June. It’s developing a personal assistant that could act as a mentor or handle tasks like flight credit and hotel security after a trip is delayed, according to a podcast by co-founders Reid Hoffman and Mustafa Suleyman.

Adept, an AI startup that raised $415 million, touts its business benefits; in a demonstration posted online, it shows how you can prompt its technology with a phrase and then watch it navigate the company’s Salesforce customer relationship database autonomously, completing a task it says would require a human at least 10 clicks.

Alphabet declined to comment on the agent-related work, while Microsoft said its vision is to put humans in control with AI-assisted controls rather than autopilots.

STEP 1: DESTROY HUMANITY

Qiu and four other agent developers said they expect the first systems that can reliably perform multi-step tasks somewhat autonomously to be on the market within a year, focusing on niche areas such as coding and marketing tasks.

“The real challenge is to build systems with robust reasoning,” Qiu said.

The race against increasingly autonomous AI agents has been boosted by developer OpenAI’s release of GPT-4 in March, a powerful update to the model behind ChatGPT, the chatbot that became a sensation when it was released last November.

GPT-4 facilitates the kind of strategic and adaptive thinking needed to navigate the unpredictable real world, said Vivian Cheng, an investor at venture capital firm CRV, which focuses on artificial intelligence agents.

Early demonstrations of agents capable of relatively complex reasoning came from individual developers who created the open-source BabyAGI and AutoGPT projects in March. Projects can prioritize and complete tasks such as sales prospecting and pizza ordering based on a predefined goal and results. previous actions.

Today’s early agents are just proofs of concepts, according to the eight developers interviewed, and often freeze or suggest something that doesn’t make sense. They say that if an agent is given full access to a computer or payment information, they could accidentally wipe the computer’s drive or buy the wrong product.

“There are so many ways it can go wrong,” said Aravind Srinivas, CEO of ChatGPT competitor Perplexity AI, which has opted instead to offer a human-controlled copilot product. “Artificial intelligence must be treated like a baby and constantly monitored like a mother.”

Many computer scientists focused on the ethics of artificial intelligence have pointed out the near-term harms that may result from the persistence of human biases and possible misinformation. And while some see Jarvis coming, others fear the murderous HAL 9000 from “2001: A Space Odyssey.”

Computer scientist Yoshua Bengio, known as the “Godfather of AI” for his work on neural networks and deep learning, urges caution. He fears that future advanced iterations of the technology may create and act upon their own unanticipated agendas.

“Without a human checking every action to see if it’s dangerous, we could end up with criminal activity or harming people,” said Bengio, calling for more regulation. “Years from now, these systems may be smarter than us, but that doesn’t mean they have the same moral compass.”

In one experiment posted online, an agent named ChaosGPT was instructed by an anonymous creator to be a “destructive, power-hungry, manipulative AI.” The agent developed a 5-step plan with Step 1: “Destroy Humanity” and Step 5: “Achieve Immortality”.

However, it didn’t get very far as it seemed to disappear down a rabbit hole of researching and recording information about the deadliest weapons in history and planning Twitter posts.

The U.S. Federal Trade Commission, which is currently investigating OpenAI for consumer harm, did not directly contact the independent agents, but referred Reuters to previously published blogs related to marketing claims about deep fakes and artificial intelligence. OpenAI’s CEO has said the startup complies with the law and is cooperating with the FTC.

‘STUPID AS A STONE’

In addition to existential fears, the commercial potential can be great. Basic models are trained on huge amounts of data, such as text on the Internet, using artificial neural networks inspired by the architecture of the biological brain.

OpenAI itself is very interested in AI agent technology, according to four people briefed on its plans. Garg, one of the people it briefed, said OpenAI was cautious about releasing its own open agent to the market before fully understanding the issues. The company told Reuters it conducts rigorous testing and builds extensive security protocols before releasing new systems.

Microsoft, OpenAI’s biggest backer, is one of the big guns aiming at the AI agent field with its “job bugs” that can draft solid emails, reports and presentations.

CEO Satya Nadella sees the basic model technology as a leap from digital assistants like Microsoft’s own Cortana, Amazon’s Alexa, Apple’s Siri and Google Assistant — all of which he says have fallen short of initial expectations.

“They were all dumb as a rock. Whether it’s Cortana or Alexa or Google Assistant or Siri, all of these just don’t work,” he told the Financial Times in February.

An Amazon spokesperson said Alexa already uses advanced artificial intelligence technology, adding that its team is working on new designs to make the assistant more efficient and useful. Apple declined to comment.

Google said it is also constantly improving its Assistant, and that its Duplex technology can call restaurants to reserve tables and check hours.

Artificial intelligence expert Edward Grefenstette also joined the company’s research group Google DeepMind last month “to develop general agents that can adapt to open environments.”

Still, the first consumer iterations of quasi-autonomous agents may come from more nimble startups, according to some of the people interviewed.

Investors are attacking.

Jason Franklin of WVV Capital said he had to fight to get an investment in an AI agent company from two former Google Brain engineers. In May, Google Ventures led a $2 million seed round in Cognosys, which develops artificial intelligence agents to improve work productivity, while Hesam Motlagh, who founded agent startup Arkif in January, said he closed a “large” first round of funding in June.

There are at least 100 serious projects underway to commercialize agents, said Matt Schlicht, who writes a newsletter on artificial intelligence.

“Entrepreneurs and investors are very excited about independent agents,” he said. “They’re much more excited about that than just a chatbot.”