Is 'AI search' without plagiarism possible?
Welcome back to Cautious Optimism! It’s Monday, June 24th, 2024. Links to hard-paywalled material will now be marked with a publication initials tag. This follows reader requests for clearer link marking. Now, AI search, the week ahead, and a final note on China’s venture capital market. To work! — Alex
C’mon Perplexity
Perplexity is a startup constantly in the news for the wrong reasons. After getting into a plagiarism spat with Forbes, Wired did some innovative reporting on the ‘AI search’ company. The publication found that it was seemingly able to get Perplexity to ignore its robots.txt file (how websites opt into the Robots Exclusion Protocol) and scrape its reporting.
Au contraire, the startup says, we’re not doing that, telling Fast Company that “Perplexity is not ignoring the Robot Exclusions Protocol and then lying about it,” but is instead leaning on third-party crawlers that might be doing the deed. Paying someone to do the bad thing instead of doing the bad thing is, in a sense, different. But not very, I would argue, and not nearly enough to get Perplexity out of the soup here.
At issue: Is Perplexity an AI search engine that uses advanced technology to answer reader questions, or is it a machine that ingests other people’s work only to serve it up half-remixed so that it can monetize research, reporting, and writing that it did not execute? More to the point, is it possible for an ‘AI search engine’ to be the former without also being the latter?
Perhaps, so long as it pays for all ingested materials. The crux here is that while it’s pretty lame for massive companies to hoover up the entire Web and use it to train AI models that they then sell without paying for the data they consumed, it’s worse when a company is ingesting data in real-time to serve up current information. Why? Because news reporting loses value over time; stealing an archive of 1880s-era New Yorker material is not great for the magazine, but taking fresh reporting from the source and selling it as your own devalues the present-day worth of reporting. Which is most of it.
Perplexity agrees that reporting has great value; after all, it’s crawling the Web and paying third parties to do the same for its product. That new information is worth a lot. Perplexity just thinks it should be able to snag and use the information for free or, perhaps, some payments later on.
There was a brief moment when media companies struck a wave of deals with AI companies. Some media folks said that it was a mistake, a trade of some short-term revenue for long-term risk. Perplexity’s actions, words, and inability to get its house in order make the skeptical voices appear more sagacious by the day.
Short Hits
Trending Up: Shopify’s reach … nuclear energy in space … parroting Russian talking points … women’s sports … water-related geopolitical tensions … extreme weather … the scrap for a slot in Apple’s AI future … Webtoon’s IPO, expected this Thursday … global shipping prices …
Trending Down: The Tories … the Yaccarino-Musk relationshio (FT) … order volume when prices are higher … India’s power grid … Apple’s attempt to meet EU regs without making less money … Nvidia (slowly) from all-time highs …
Coming Up: May consumer confidence on Tuesday, new home sales Wednesday, initial jobless claims and a new Q1 GDP number Thursday along with durable-goods orders and pending home sales. Friday brings a wealth of information, more here.
Earnings This Week: We’ll hear from FedEx on Tuesday, Micron Technology Wednesday, and Nike on Thursday. Q2 cometh.
The Markets: Asian stocks were largely lower today, while shares in Europe generally rose. Ahead of the bell, domestic shares are mixed. Crypto prices are falling.
China’s Venture Exit Problem
China’s venture capital market is struggling in the post-COVID period, with investment totals falling and venture funds’ ability to raise capital decelerating. Recent reporting indicates that the country’s exit market could be even more frozen than what we see in the United States and Europe.
To put that into numerical terms, KPMG data reports that Chinese companies raised $15 billion in Q4 2023 and $11.5 billion in Q1 2024, out of regional totals of $22.9 billion and $18.9 billion across the same timeframe.
As we discussed last week, the reasons for the slowdown are several. Government regulation, government meddling, geopolitical tensions, and a trade war all play a part. One reason venture investment in China is weak, however, is incredibly pedestrian: exits.
IPO activity in China has fallen off a cliff. Here’s Nikkei Asia:
During the first six months of the year, the total value of mainland China's IPOs has plummeted 84% on the year to 32.5 billion yuan ($4.48 billion), while only 44 companies went public, down 75%. […] While Hong Kong's market has been drying up for a while, the shift on the mainland came after China's securities watchdog introduced stricter regulations in mid-March to shore up market confidence. This has contributed to a significant decline in IPO applications, with only two companies applying in the past six months, compared with 334 last year.
The article notes that Hong Kong’s IPO market has been in the dumps since 2021, unsurprisingly.
Fewer Chinese exits mean fewer venture capital dollars flowing into Chinese startups. But there’s another, further upstream impact: venture firm fundraising. Nikkei Asia reports Dealogic data indicating that Chinese VCs raised two-thirds less in 2023 than the year before ($14.1 billion) and that the trend has continued into this year with a “30% year-on-year decline as of May.”
It’s going to be hard to bring back 1. Venture firm fundraising if 2. Exits are limited, which means that 3. Startups in the country are going to be capital-starved for some time.
China’s tech market is broken from the source (LPs) to the finish line (exits). It will take a while to fix if that is even possible under the current government.
The second quarter is coming to an end. That means Q2 venture capital data and Q2 earnings are just around the corner. Make sure you are subscribed to Cautious Optimism, because we have a lot of work to do. Chat soon! — Alex