Therese Poletti
As the tech world goes wild with the latest chatbots, remember Watson’s AI fame and subsequent influence before getting too excited about the possibilities of the latest machine learning genius.
ChatGPT is the latest in artificial intelligence to breathe (and venture capital investment into) Silicon Valley, but it’s also an advance in AI that hasn’t proven its ability to live up to the hype.
ChatGPT debuted about a month ago and offers a much better chatbot than its predecessor. He said OpenAI, the company that created ChatGPT, is currently in research preview and is collecting data to train so-called large language models.
Projects like ChatGPT can be amazing because they test the limits of technology and try to push it further. However, the anticipation of technological possibilities does not always lead to great financial gains.
Take Watson as an example. IBM Corp. (IBM)’s natural language computer famously beat two humans on the game show Jeopardy more than a decade ago, prompting Big Blue to bet big on the technology. became. IBM has opened a business unit with his IBM Watson offices on both the East and West Coasts, focusing on healthcare as the biggest business opportunity.
That opportunity did not lead to IBM’s wealth. Healthcare’s push has disappointing, and when the company streamlined some of its operations earlier this year, it sold Watson Health assets to a private equity firm for an undisclosed amount. is a suite of AI tools and apps in its $25 billion software business, but IBM doesn’t disclose how much revenue these products generate. IBM’s data/AI segment revenue declined 1% year-over-year in the third quarter. In February 2011, IBM’s stock fell 8.6% since Watson won “Jeopardy,” according to FactSet data, while the S&P 500 index rose 189.5% in the meantime. .
That lesson hasn’t stopped the fascination with AI development. Forrester Research analyst Rowan Curran told his MarketWatch that ChatGPT was “his AI moment for the iPhone,” and in 2007 he noted that Apple Inc. (AAPL) reinvented consumer electronics. said he did. “This is a big deal,” he added, adding that this is where technology is headed.
Don’t Miss: Opinions: Delivery Drones, Robo-Taxis and Even Insurance — The much-touted dream for AI startups is a nightmare for tech investors
OpenAI is private and does not disclose earnings, but valuations and investments are rising rapidly. The San Francisco company was founded as a non-profit in 2015, and its first press release at the time featured Silicon Valley investors such as Peter Thiel and Reed Hoffman and Amazon.com Inc.’s (AMZN) Amazon Web Services, among others. , the valuation has not been disclosed. Elon Musk was also co-chairman, and Sam Altman is still chief executive officer (Musk stepped down in 2018 after his resignation, citing potential conflicts of interest with his AI business at Tesla (TSLA). resigned for the reason).
A lot has changed since then — OpenAI raised $1 billion in funding from Microsoft Corp. (MSFT) alone in 2019. Microsoft is also reportedly looking to pour more money into a “marginal margin” company that was recently valued at about $20 billion in a year. Stock sale. In a recent pitch to investors, an OpenAI executive said he expects revenue to be around $200 million next year, and he expects to hit $1 billion by 2024, according to a Reuters report. .
OpenAI now offers ChatGPT’s underlying multilingual model, GPT-3, as a paid application programming interface (API) service, so businesses can build apps on top of it. Reuters reported that OpenAI charges around 1 cent or more for text-based queries and 2 cents for image results.
According to IDC analyst Ritu Jyoti, some ChatGPT customers are using GPT-3-powered chatbot negotiators to help consumers cancel subscriptions, file parking tickets, and generally That includes DoNotPay, a startup that helps fight big companies. He has one customer, Jasper, who uses GPT-3 to power his AI-based copywriting and content creation software.
More: What is ChatGPT? Well, you can ask it yourself
OpenAI officials did not respond to requests for interviews, so MarketWatch turned to the free version of ChatGPT available on the web. When MarketWatch asked ChatGPT, “Who are the customers of GPT-3?” Then I listed some customers, including eBay Inc. (EBAY). The company says he uses GPT-3 to improve product classification and search accuracy. Goldman Sachs Group Inc. (GS) uses it to analyze large amounts of unstructured data. The Guardian uses it to generate news articles.
However, these answers may not be the most accurate, but they fit the current state of ChatGPT despite the excitement. Goldman Sachs was part of the GPT-3 demo, stating that it uses machine learning and other AI techniques to infer information, while Guardian only wrote using GPT-3. A few pointless paragraphs in an AI article earlier this year that won’t lead to long-term revenue.
“It’s unimaginable what it can do,” said Jyoti, adding that he was “thrilled and impressed” by OpenAI’s ChatGPT demo. “But I live and breathe in this space, so I want to be aware that there are a lot of inaccuracies, they don’t tell you where the data comes from, and they can be very misleading.”
OpenAI CEO Altman said of the limitations of chatbots:
“ChatGPT sometimes writes answers that sound plausible but are inaccurate or nonsensical,” OpenAI said in a blog post about the release.
Saffo said he isn’t too concerned about the chatbot’s propensity for errors. He’s “certainly more accurate than your average conspiracy theorist,” he half-jokingly said. “This is a demonstration of the technology and the first demo to come.”
One of the confirmed customers MarketWatch spoke with is Smartling, a cloud-based smart translation service company. Smartling uses her GPT-3 to improve the source text for translation in a post-editing process before the text is passed to translation.
Olga Beregovaya, Vice President of AI and Machine Translation at Smartling, said:
Forrester’s Curran says, “Just because you’re generating coherent nonsense doesn’t mean your content isn’t useful.” “ChatGPT is a very good model. Generally, it’s great at explaining why certain things can’t be done…if you’re using ChatGPT for internal customer service representatives, The question to be asked is: If the range is much narrower, the training data will be known.”
Details from Therese: Unprofitable public company bets over $200 million in California, suffers huge losses
However, the use cases are so narrow that it also narrows the revenue potential, but there is not much about AI yet. Global spending on AI, including software, hardware and services for AI-centric systems, will reach nearly $118 billion in 2022 and exceed $300 billion in 2026, according to IDC. , market research firms have not pioneered the market for AI software. Unique. And considering how expensive top-of-the-line hardware for AI is, software is likely a small percentage of the total.
So it’s hard to know just how big or small these AI software companies are today, and what the future holds. As many analysts have pointed out, it’s definitely still early days. But even if that were the case, Mr. Saffo said, “That’s what we’re aiming for.” He also said the excitement surrounding tech demos throughout and his ChatGPT shows that humans always want to believe that inanimate objects have intelligence.
“The relentless pace of Moore’s Law means that these things go on in doubling periods of 18 to 20 months,” he said. “People will look back at ChatGPT and say, ‘It was so basic and boring.'”
The same can be said for Watson ten years later, which made a big splash. And I would say that the ability to make money was even more disappointing than its endurance as a technology. Will he be able to say the same about ChatGPT in his next decade?
– Therese Poletti
(Closed) Dow Jones Newswire
12-27-22 1940ET
Copyright (c) 2022 Dow Jones & Company, Inc.
