About a decade after virtual assistants like Siri and Alexa boomed, a new wave of AI assistants with greater autonomy is up the stakes, powered by the latest version of the technology behind it. after ChatGPT and its competitors.
Test systems running on GPT-4 or similar models are attracting billions of dollars in investment as Silicon Valley compete to take advantage of advances in AI. The new assistants – often referred to as “agents” or “co-pilots” – promise to perform more complex work and personal tasks when commanded by humans without close supervision.
“Senior, we want this to be something like your personal AI friend,” said developer Div Garg, whose company MultiOn is beta testing an AI agent.
“It could grow into Jarvis, where we want this to be connected to many of your services,” he added, referring to Tony Stark’s integral AI in the Iron Man movies. “If you want to do something, you talk to your AI and it will do your thing.”
The industry is still far from mimicking the brilliant digital assistants of science fiction; For instance, Garg agents browse the web to order burgers on DoorDash, while others can create investment strategies, email refrigerator sellers on Craigslist, or summarize business meetings for late participants.
“A lot of things that are easy for humans are still extremely difficult for computers,” said Kanjun Qiu, CEO of General Intelligent, an OpenAI competitor creating AI for agents.
“Let’s say your boss needs you to schedule a meeting with an important group of clients. That involves complex reasoning skills for AI – it needs to understand people’s preferences, resolve conflicts conflicts, while maintaining the necessary care when dealing with clients.”
According to Reuters interviews with about two dozen entrepreneurs, investors and AI experts.
The new technology has caused a craze towards assistants powered by the so-called GPT-4 inclusion platform model, attracting individual developers, celebrities such as Microsoft and Google parents Alphabet plus a bunch of startups.
Inflection AI, a startup, raised $1.3 billion at the end of June. It is developing a personal assistant that it says can act as a mentor or handle tasks services such as securing flight and hotel credits after travel delays, according to a audio files by co-founders Reid Hoffman and Mustafa Suleyman.
Adept, an AI startup that has raised $415 million, advertises its business interests; in a demo posted online, it shows how you can prompt its technology with a sentence, then watch it navigate the company Sales force its own customer-relational database, completing a task it said would require 10 or more human clicks.
Alphabet declined to comment on agency-related work, while Microsoft says its vision is to keep humans in control of the AI co-pilots, rather than autopilot.
STEP 1: KILL HUMANITY
Qiu and four other agency developers say they expect the first systems that can reliably perform multi-step tasks with some autonomy to hit the market within a year, focus on narrow areas like coding and marketing tasks.
“The real challenge is building systems with solid reasoning,” says Qiu.
The race towards increasingly autonomous AI agents was spurred by the release of GPT-4 in March by developer OpenAI, a powerful upgrade of the model behind ChatGPT – the chatbot that has become phenomenon when it was released last November.
Vivian Cheng, an investor at venture capital firm CRV who focuses on AI agents, says GPT-4 facilitates the kind of strategic thinking and adaptability needed to navigate in the unpredictable real world.
Early demonstrations of relatively complex reasoning agents came from individual developers who created the BabyAGI and AutoGPT open source projects in March, which can prioritize and execute effects services like prospecting and ordering pizza based on predefined goals and the results of previous actions.
According to eight developers interviewed, today’s original resellers are merely proof of concept, and often freeze or suggest something nonsensical. If given full access to a computer or payment information, a dealer could accidentally erase a computer’s drive or buy the wrong thing, they said.
“There are so many ways it can go wrong,” said Aravind Srinivas, CEO of ChatGPT competitor Perplexity AI, who chose to offer a human-supervised copilot product. . “You have to treat the AI like a child and constantly supervise it like a mother.”
Many computer scientists who focus on AI ethics have pointed out the short-term harm that can come from perpetuating human biases and the potential for misinformation. And while some see a future Jarvis, others fear the murderous HAL 9000 from “2001: A Space Odyssey”.
Computer scientist Yoshua Bengio, dubbed the “godfather of AI” for his work on neural networks and deep learning, calls for caution. He fears future cutting-edge iterations of the technology could create and act on their own unexpected goals.
“Without someone checking every action to see if it’s dangerous, we could end up committing crimes or potentially harming people,” Bengio said, calling for a more regulations. “In many years from now, these systems could be smarter webut that doesn’t mean they have the same moral compass.”
In an experiment posted online, an anonymous creator instructed an agent called ChaosGPT to become a “destructive, power-hungry and manipulative AI”. The Agent has developed a 5-step plan, with Step 1: “Destroy Humanity” and Step 5: “Achieving Immortality”.
It didn’t go too far, though, seemingly disappearing into a rabbit hole that researched and archived history’s deadliest weapons and scheduled Twitter posts.
The US Federal Trade Commission, which is currently investigating OpenAI over concerns about consumer harm, did not mention autonomous dealers directly, but referred Reuters to pre-published blogs. here about deepfakes and marketing claims about AI. The CEO of OpenAI said the startup complies with the law and will work with the FTC.
‘SONE PEOPLE’
Existential fears aside, the commercial potential can be enormous. Platform models are trained on large amounts of data such as text from Internet uses an artificial neural network inspired by the architecture of biological brains.
OpenAI itself is very interested in AI agent technology, according to four people briefed on its plans. Garg, one of those briefed on the report, said OpenAI is wary of releasing its own open agent to the market before fully understanding the issues. The company told Reuters it conducts rigorous testing and builds extensive safety protocols before releasing new systems.
Microsoft, OpenAI’s biggest backer, is one of the big guns that is targeting the AI agent field with its “co-pilot for the job” that can draft solid emails, reports, and presentations.
CEO Satya Nadella consider platform modeling technology as a leap forward from digital assistants like Microsoft’s Cortana, Amazon’s AlexaApple Siri and Google Assistant – which, according to him, all fell short of initial expectations.
“They’re all rock-stupid. Whether it’s Cortana or Alexa or Google Assistant or Siri, all of these things don’t work,” he told the Financial Times in February.
ONE Amazon The spokesperson said that Alexa has used advanced AI technology, adding that their team is working on new models that make the assistant more capable and useful. Apple declined to comment.
Google says it’s also constantly improving its assistant, and that its Duplex technology can call restaurants to reserve tables and verify hours.
AI expert Edward Grefenstette also joined Google’s DeepMind research team last month to “develop generic agents that can adapt to open-ended environments.”
However, according to several interviewees, the first consumers to use quasi-autonomous agents may come from more nimble startups.
Investors are piling up.
WVV Capital’s Jason Franklin says he struggled to get an AI-agents investment from two former Google Brain engineers. In May, Google Ventures led a $2 million funding round in Cognosys, developing AI agents to boost productivity, while Hesam Motlagh, who founded agency startup Arkifi, in January, said he closed a “fairly large” first round of funding in June.
Matt Schlicht, who writes an AI newsletter, says there are at least 100 serious projects working on commercializing agents.
“Businessmen and investors are extremely excited about self-driving agents,” he said. “They’re more excited about it than simply about a chatbot.”
