Nearly a decade after virtual assistants fell in love Sir, I And Alexa burst onto the scene, a new wave oh Helpers with more autonomy are raising the stakes powered by the latest version of the technology behind chatgpt and its rivals.
experimental systems that run GPT-4 Or similar models are attracting billions of dollars in investment as Silicon Valley competes to capitalize on advances in AI. The new assistants – often called “agents” or “copilots” – promise to perform more complex personal and task tasks when commanded by a human, without the need for close supervision.
“At the high level, we want it to be something like your personal AI friend,” said developer Div Garg, whose company MultiOne is beta-testing an AI agent.
“It could evolve into Jarvis, where we want it to be connected to a lot of your services,” he said, referring to Tony Stark’s indispensable AI in the Iron Man movies. “If you want something done, you talk to your AI and it does things for you.”
The industry is still a long way from emulating the dazzling digital assistants of science fiction; For example, Garg’s agent browses the web to order burgers on DoorDash, while others may create investment strategies, email people selling refrigerators on Craigslist or summarize work meetings for latecomers. can give.
“A lot of what’s easy for people is still incredibly difficult for computers,” said Kangjun Qiu, CEO of Generally Intelligent. OpenAI Creating AI for competitive agents.
“Let’s say your boss needs to schedule a meeting with a group of important customers. This involves reasoning skills that are complex for the AI – while maintaining the careful touch required when dealing with customers, everyone needs to be achieved, conflicts resolved.”
The initial efforts are only a taste of the sophistication that could come from increasingly advanced and autonomous agents in future years as the industry moves toward an artificial general intelligence (AGI) that could equal or surpass humans in a myriad of cognitive tasks. is, as stated in the interview with Reuters. Two dozen entrepreneurs, investors and AI experts.
The new technology has led to a rush toward assistants powered by the so-called Foundation model, including the GPT-4, which has led to the involvement of individual developers, such as big hitters. Microsoft And Google Parents Alphabet Also a lot of startups.
To name one startup, Inflection AI raised $1.3 billion (roughly Rs. 10,663 crores) at the end of June. According to a podcast by co-founders Reed Hoffman and Mustafa Suleiman, it’s developing a personal assistant that it says can act as an advisor or perform tasks such as securing flight credits and hotels after travel delays. can handle.
Adept, an AI startup that has raised $415 million (roughly Rs. 3,404 crores), talks about its business benefits; In a demo posted online, it shows how you can prompt its technology with a single sentence, and then watch it automatically navigate the company’s Salesforce customer-relationship database, completing a task. It states that a human would need 10 or more clicks.
Alphabet declined to comment on the agent-related work, while Microsoft said its approach is to put humans in control of AI copilots, rather than autopilots.
Phase 1: Destroy Humanity
Qiu and four other Agent developers said they expect within a year the first systems will be on the market that can reliably execute multi-step tasks with some autonomy, focusing on narrow areas such as coding and marketing tasks. Will be focused
“The real challenge is to build systems with strong logic,” Qiu said.
The race toward increasingly autonomous AI agents has been supercharged by developer OpenAI’s March release of GPT-4, a powerful upgrade to the model behind ChatGPT – the chatbot that became a sensation when it was released last November.
GPT-4 facilitates the strategic and adaptive thinking needed to navigate the unpredictable real world, said Vivian Cheng, an investor at the venture capital firm CRV, which focuses on AI agents.
Early demonstrations of agents capable of comparably complex reasoning came from individual developers, who in March created the BabyAGI and AutoGPT open-source projects that prioritize tasks such as sales prospecting and pizza ordering based on pre-defined objectives and outcomes. can and execute. past actions.
According to eight developers interviewed, today’s early crop of agents are only proofs of concepts, and often end up withholding or suggesting something that doesn’t make sense. They say that if an agent is given full access to a computer or payment information, he could accidentally erase a computer drive or buy the wrong thing.
“There are so many ways this could go wrong,” said Arvind Srinivas, CEO of ChatGPT competitor Perplexity AI, which has opted to introduce a human-supervised co-pilot product. “You have to treat AI like a child and constantly monitor it like a mother.”
Many computer scientists focused on AI ethics have pointed to the near-term harm that can result from the persistence of human biases and the potential for misinformation. And while some see the Jarvis of the future, 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. They fear that future advanced iterations of the technology may make and act upon their unexpected targets.
Bengio called for more regulation, saying, “Without someone who checks every activity to see if it’s not dangerous, we may be engaging in actions that are criminal or could harm people. ” “Years from now these systems may be smarter than us, but that doesn’t mean they have the same ethical guidelines.”
In one experiment posted online, an anonymous creator instructed an agent named ChaosGPT to become a “destructive, power-hungry, manipulative AI”. The agent developed a 5-step plan, Step 1: “Destroy Humanity” and Step 5: “Obtain Immortality”.
It didn’t go far enough, however, to disappear down a rabbit hole of researching and storing information and planning Twitter posts about history’s deadliest weapons.
The US Federal Trade Commission, which is currently investigating OpenAI over concerns of consumer harm, did not directly address autonomous agents, but referred Reuters to previously published blogs on deepfakes and marketing claims about AI. OpenAI’s CEO has said that the startup abides by the law and will work with it FTC,
‘Dumb as a rock’
Existential fears aside, the business potential could be huge. Foundation models are trained on vast amounts of data such as text from the Internet using artificial neural networks that are inspired by the architecture of the biological brain.
OpenAI itself is very interested in AI agent technology, according to four people with knowledge of its plans. Garg, one of the people briefed on this, said OpenAI is wary of releasing its own open-ended agent to market before the issues are fully understood. The company told Reuters it conducts rigorous testing and has extensive security protocols in place before releasing new systems.
Microsoft, OpenAI’s biggest supporter, is one of the big guns taking aim in the AI agent field with its “co-pilot for work” that can draft solid emails, reports and presentations.
CEO Satya Nadella Foundation-model sees the technology as a leap forward from Microsoft’s own digital assistants Cortanaamazon alexa, of apple Siri and google assistant – In his view, all of these fall short of initial expectations.
“They were all as dumb as a rock. Whether it’s Cortana or Alexa or Google Assistant or Siri, they all don’t work,” he told the Financial Times in February.
An Amazon spokesperson said that Alexa already uses advanced AI technology, adding that its team is working on new models that will make the assistant more capable and useful. Apple declined to comment.
Google said it is also constantly improving its Assistant and that its Duplex technology can call restaurants to book tables and verify hours.
AI expert Edward Grefenstedt also joined Google DeepMind, the company’s research group, last month to “develop generic agents that can adapt to open environments”.
Still, the first consumer iteration of semi-autonomous agents could come from more nimble startups, according to some of the people interviewed.
investors are jumping in
Jason Franklin of WVV Capital said he had to wrestle two former Google Brain engineers to invest in the AI-agent company. In May, Google Ventures led a $2 million seed round in Cognosis, which develops AI agents for work productivity, while in January Hessam Motlagh, who founded agent startup Arcify, said he Has completed the “big” first financing. round in June.
Matt Schlicht, who writes a newsletter on AI, said there are at least 100 serious projects working to commercialize agents.
“Entrepreneurs and investors are very excited about autonomous agents,” he said. “They’re far more excited about it than any chatbot.”
© Thomson Reuters 2023