Slashdot Mirror
As AI Explodes, Investors Pour Big Bucks Into Startups (siliconangle.com)
Investment in AI startups is on a tear as venture capitalists and corporate investors scramble to stake out a leadership position in what could be the driving trend in technology for decades to come. From a report: The financial interest in AI, machine learning and related technologies is hardly new. CB Insights has tracked some $18.4 billion invested in 2,541 AI-related startups since 2012. But the trend is only accelerating. In the latest MoneyTree report from PricewaterhouseCoopers and CB Insights, which showed otherwise mostly stagnant startup funding, AI and machine learning companies shined, reaching an eight-quarter high of $820 million invested in 90 companies. A flurry of significant investments in a number of AI-related companies this past week underscored the point. On Wednesday alone, for instance, AI-powered analytics software provider CognitiveScale raised a $15 million round, voice AI startup Snips raised $13 million and, to top it off, machine learning consultancy Element AI got an unusually large $102 million early-stage investment just eight months after the company was launched. Then on Thursday and Friday, two other AI-powered companies, Conviva and Codota, announced fundings too.
Could this be the foreshadowing of another AI Winter? I remember the AI Winter of the late 1980's and early 1990's. At that time, the hype was about some of the truly amazing things that could be done in Prolog like languages. Pattern matching. Deductive reasoning. Theorem provers. Computer Algebra Systems (CAS). And especially Expert Systems.
The expectations got totally out of control. Wow! A knowledge expert could write a set of rules so that an expert system could predict who is a bad credit risk! Etc. Of course, modern statistical approaches might be much better at that. But I use it as an example of having too great of expectations.
Like today, these modern statistical classifiers are amazing! But one day one of those statistical classifiers will mis-classify a pedestrian in front of a vehicle. Another possible way there could be wrong expectations is that both human beings and also managers might expect these systems to have some kind of insight or creativity. Or possibly deductive reasoning power (like the classic AI systems actually had, to a degree).
I'll see your senator, and I'll raise you two judges.
There have been gradual speed-improvements, but they are, well, gradual.
Even with gradual speed improvements, you may reach a speed that represents a tipping point, where these processes go from taking an unacceptably long amount of time to taking an acceptably long amount of time.
When you hit a tipping point like that, usage and adoption might expand dramatically (even if not "explosively), regardless of whether there's been real innovation.