New research is highlighting a major problem ahead for the artificial intelligence industry, just as sector superstar Nvidia’s upcoming earnings are set to test investors’ strained relationship with tech stocks. Today, we look at why market sentiment is glitching – and what AI giants might do to fix it.
Up top
Corus vs. Rogers: Corus Entertainment Inc. is arguing that Rogers Communications Inc. has used its dominant industry position to unfairly disadvantage Corus and other independent media companies, according to a complaint filed with Canada’s broadcasting regulator.
Taking on the tariff: Canadian lumber industry and government officials say they will defend the country’s economic interests as they seek a resolution after the U.S. Department of Commerce nearly doubled its tariffs on softwood this week.
Capital gain pains: Ottawa’s efforts to blunt the impact of its recent capital-gains tax changes on small business owners and entrepreneurs don’t go far enough, with many still facing a higher tax burden when they sell their companies, business groups say.
In focus
AI’s next task: delivering results
I love stock photos so much.
Globe reporter Joe Castaldo wrote this week about new research that shows AI models “collapse” and spout gibberish over time. A fix could eventually emerge, but the findings come as companies spend big and investors recalibrate their tech enthusiasm against a fuller economic backdrop. I asked Castaldo to help make sense of a pivotal moment for the AI industry.
What is generative AI?
When people talk about generative AI, they really mean AI applications that produce text, images, video and audio. The most high-profile example is OpenAI’s chatbot, ChatGPT, which came out almost two years ago.
These underlying models are built on a ridiculous amount of data to find patterns, allowing them to predict the next word in a sentence, for example. ChatGPT was likely the first time a lot of people had direct exposure to AI.
Where is it used now?
Lots of places – predicting weather patterns, fraud detection, forecasting when industrial machines will need servicing – but all of that is pretty abstract. With applications like ChatGPT, suddenly everyone could try it out themselves. And that got a lot of people very excited.
And in the corporate sphere?
In the business world, companies are using generative AI for customer service, such as by building chatbots to interact directly with customers or to help call centre agents find answers faster.
There are startups using generative AI to help lawyers answer legal questions, find case law, write contracts, and so on. The same thing is happening in human-resources departments.
There are companies selling AI software to medical offices to transcribe patient consultations and generate summaries, so doctors don’t have to take notes while simultaneously carrying on conversations and diagnosing ailments. And copywriting press releases and marketing materials is also a big area.
What are the biggest promises being made by companies at the forefront of the technology and its applications?
Some developers are pursuing artificial general intelligence, meaning AI models that are much smarter than we are. Depending on who you ask, that could be a very good thing for humanity or usher in a dystopia where we’re ruled by robot overlords. More practically, companies like Canada’s Cohere just want to use generative AI to help businesses solve problems. Broadly, the big promise of AI is efficiency: automating a lot of tasks, boosting productivity, saving money and liberating employees to tackle more complex things that AI isn’t as good at today.
Have any of those promises either been broken or stalled?
There is a lot of skepticism around generative AI, but there are some areas where it’s already proven useful. Companies that I’ve talked to say coding with generative AI, though far from perfect, can nevertheless save time and allow engineers to be more productive.
In other areas, the true impact is less clear. Generative AI can make mistakes and invent information, which is called hallucinating. It’s hard for a business to put out a product without certainty that it’s going to work properly. Costs can be high, and businesses have to figure out what the return on investment will be. If I work in marketing and use generative AI to help me draft a press release, how much time am I spending crafting the right prompt and then re-writing whatever the application spits out? Would it have been faster and higher quality if I’d written it from scratch?
Why has the market lost some of its enthusiasm over the last couple of months?
Businesses and consumers are definitely more aware of the limitations of generative AI after the “wow” factor of ChatGPT passed. Corporations have lots to consider, such as data privacy and security, so it’s not necessarily straightforward to implement this stuff. Other companies are proceeding very cautiously, too. To give an indication of how slowly some corporations are moving, TD Bank put out a press release this year to say it had completed a pilot project with an AI-coding assistant.
And then there’s cost. Building generative AI models is enormously expensive, and companies like Microsoft, Amazon and Google are spending billions on data centres to power AI computing.
What is the biggest hurdle for AI companies to overcome if they want to placate investors and clients?
They have to show results: How is generative AI saving money? Improving productivity? Unlocking new ideas? And they need to convince people to be patient. ChatGPT hasn’t even been out for two years yet, so it’s early days. At least one big test will be when the next iteration of the really big frontier AI models come out, like the successor to OpenAI’s GPT-4. How much better will it be? That might provide clarity on where this is all heading.
Mailbag
Banking on a stock split?
A Royal Bank of Canada on Bay Street.Reuters
Yesterday, we opened the floor to questions. Today, reader Dave Abbott in Austin, Tex., asks:
“What are the odds, in your opinion, that RBC will have a 2-for-1 split in the near term? CIBC had a 2-for-1 split approximately a year and a half ago and the split price has been trending up.”
I put the question to veteran investment reporter David Berman, whose outlook on bank stocks and strategy comes from a far deeper well of experience than mine:
“Bank CEOs don’t care much about stock splits these days. Through electronic trading, investors are no longer limited to buying board lots of 100 shares, which means that the nominal share price has little meaning.
As well, the argument that share splits drive share prices higher is a bit flimsy. Although there is some evidence suggesting that stocks perform well after a split, companies generally make the decision to split their shares only when they’re on a winning streak. So, further gains are not likely driven by the split itself.
Bank earnings season, which kicks off next week, might provide some indication of where the banks now stand on the issue. The odds of RBC splitting its shares over the long-term are high. In the near-term, though, investors shouldn’t count on it – or care about it.”
Have a question or a story idea? Email me: cws@globeandmail.com
The outlook
Today: A weighty mix of earnings including Walmart, Alibaba, Deere & Co. and Macy’s. Canada reports wholesale trade and new motor vehicle sales for June, and existing home sales for July.
Next week: On Tuesday, Statistics Canada reports the consumer price index for July. The Bank of Canada’s next interest rate announcement is on Sept. 4.
In space: NASA has once again delayed its decision on when the troubled Boeing Starliner spacecraft can leave the International Space Station.
Spaced out: The Globe’s Barry Hertz gets the top two titles in the wrong order in his ranking of Alien movies.
Morning markets
Global markets were on firmer ground, continuing to recover from the early August rout on optimism that the U.S. Federal Reserve would start to cut interest rates soon. Wall Street futures and TSX futures were pointing higher.
Overseas, the pan-European STOXX 600 was up 0.09 per cent in morning trading. Britain’s FTSE 100 rose 0.02 per cent, Germany’s DAX advanced 0.23 per cent and France’s CAC 40 declined 0.07 per cent.
In Asia, Japan’s Nikkei closed 0.78 per cent higher, while Hong Kong’s Hang Seng slipped 0.025 per cent.
The Canadian dollar traded at 72.94 U.S. cents.