That's the conventional argument for government funded research, but is it true? The AI space is a good example of very long term research, funded almost entirely by the private sector without any clear idea of how the results would be marketed, they publish and - the key part - the work actually replicates. Nor is the AI space unique. XEROX Parc was another famous example, Big Data/Cloud has come entirely out of private sector research. Not unique to computer science either: the research work that led to Ozempic was being funded by Novo Nordisk as far back as 1998.
> Research usually doesn't end in something that's marketable
It usually does, even in academia. That's why they count citations. It's just that the thing that's marketed is either policy advocacy (marketed to governments) or the claims in and of themselves, marketed either to the general public in TED talk style advocacy or to other academics as work they can build on to get more grants. A lot of people don't recognize the existence of things like the profit motive, lobbying or marketing in government funded research, imagining that they don't exist, but they do. In some fields, it's common for lobbying to be like 50% or more of the word count.
There's a long tail of papers that research things that nearly nobody cares about, yet perhaps one day they might. These are the "we studied some obscure fish in the Amazon and found a cure for cancer" type stories that crop up from time to time. But such claims often don't quite work out, and the private sector is easily able to fund this sort of exploratory work too.
Yet at the same time we keep seeing startups bringing innovation, while the big enterprises do not bring that much original innovation. Instead, those big enterprises rather kill innovation by buying those startups. Not seldom to just cancel the whole product together.
This is almost a secret, but __most innovation come from the small players__.
Those small players benefit from open access and academic progress. It is a very fruitful cooperation.
But the tragedy is that in the US the meaning of "the free market" is being loudly distorted.
What they actually mean is a cozy climate for oligarchs, protected by the mountains of tariffs and deregulation. In the US society, the idea of competition has deformed into "killing competition", and they clarify it now with "by all means possible". Those extremists HATE competition. Nobody arrived likes being disturbed and having to work to keep competitive. Rent extraction is way simpler.
Especially HN should think a little longer about what that all means for aspiring entrepreneurs.
That is not actually what happens with most acquisitions, especially not in the drug and biotech space. Big enterprises buy promising startups at the phase 1 or 2 stage fully intending to bring them to market. But still many fail to ever make it through phase 3 and gain FDA approval. That's not an innovation problem it's just the nature of the business.
I am not sure if this applies here, but Regulatory Capture is a certain way to block the small players from growing. I understand that drugs are a special case, so I do not know if that applies there.
Also, the fact that a product fails post-acquisition is precisely what happens a lot in regular business too. They can be seen as a threat to other fiefdoms in the behemoth, or they lose their autonomy. Doesn't always mean the product or idea was flawed beyond repair.
Well, in the AI space at least that's not really true. Google, NVIDIA and Meta have all made huge contributors. So has Microsoft, in their own way. It's a real mix of big old companies and smaller upstarts.
Even in this example it is a yes and no. Remove academic development in the AI, including the paths we left, and your examples would not have any AI at all.
The problem is a little bit more nuanced than that though. It is not that big companies do not make inventions. Often, these inventions do not survive politically in the behemoth. That is why some companies decide to create a spinoff, knowing that their own body would be trying to kill the growth of their offspring.
Even in the early days AI was heavily funded by the private sector. Symbolics machines were partly promoted as a way to do AI research, back in the logic era. And the AI winter was mostly a grant funding phenomenon. When it became unfashionable in universities the field rebranded as ML and became commercially driven by (mostly) different people. Companies like Google invested in advanced ML research from day one.
Nowadays it's been rebranded back to AI due to the switch to neural methods, but there's been funding for AI from the computer industry for as long as the field existed.
I'm pretty sure AI would exist as a field and be in a similar place to where it is now, even if governments had never funded it at all.
Even more, a significant portion of the researchers at industrial labs got their start as graduate students, largely funded by government grants. Even if somehow no actual fundamental research transferred from academia into industry, the people sure do.
> Research usually doesn't end in something that's marketable
It usually does, even in academia. That's why they count citations. It's just that the thing that's marketed is either policy advocacy (marketed to governments) or the claims in and of themselves, marketed either to the general public in TED talk style advocacy or to other academics as work they can build on to get more grants. A lot of people don't recognize the existence of things like the profit motive, lobbying or marketing in government funded research, imagining that they don't exist, but they do. In some fields, it's common for lobbying to be like 50% or more of the word count.
There's a long tail of papers that research things that nearly nobody cares about, yet perhaps one day they might. These are the "we studied some obscure fish in the Amazon and found a cure for cancer" type stories that crop up from time to time. But such claims often don't quite work out, and the private sector is easily able to fund this sort of exploratory work too.