A well-balanced research portfolio ought to include long shots. Sometimes they come up big, big winners. But they spend a lot of time looking weird and pointless.
Unfortunately, funding long-shots is really hard. I was going to write "in the present environment", but then I realized that was too specific. It doesn't really matter where the money is coming from. Funding long shots is fundamentally hard, whether it's from private investment, government, commercial, or whatever else you can think of. But it's still a good idea to put a small portion of the research resources into those long shots.
(Now, it may be the case that the Media Lab is a bad place to fund your longshot, but I don't know enough to evaluate that.)
Some organisations fund internal research so that they can demonstrate innovation to customers / investors, etc during marketing / sales, e.g. to differentiate themselves from competitors. If some of the ideas are in fact workable / profitable that is a bonus.
My first job in industry outside academia was as a programmer in the 'Advanced Techniques Department' (a.k.a. Toy Shop) of a traditional UK engineering company. We were quite often incorporated into the agenda of senior VIP visits, and we were instructed to focus on the innovative aspects of what we did. I realised later that the demos were actually great training for me as well, giving me insights into communicating with non-technical executives. I also got to work with European partner research groups in industry (spells in Denmark and Germany), which was also excellent experience.
Incidentally, the ATD had much better tooling for AI research (Lisp and a small fleet of Lisp machines) than I'd used for research at Uni (C / Unix). I was also introduced to proper software engineering discipline. My productivity (in terms of demo'd systems / capabilities) was much higher than it had been as a postgrad, which I quickly came to realise had been undisciplined hacking.
A primary role for corporate R&D is to wow investors and competitors via splashy sexy demos that your company is not only cutting edge but is actually reinventing the business. This is true for the media, communication, art, HCI, and architecture industries as well as science- or engineering-based ones. The Media Lab was a venue to promote those kinds of innovations in ways that papers in research journals never could.
The Media Lab brought together a diversity of nontraditional tech interests and capabilities in one place to explore how cutting edge ideas in how we interact with information might disrupt everyday life in artful ways, creating cool demos to titillate the masses and foment fear in the hearts of competitors.
And it largely worked, for the first 15 years or so, anyway. But eventually, Being Digital (or WiReD) wasn't all that disruptive or cool any more, and the bloom was off the rose.
To change the game, to change the way people think about technology and the context of technology, to understand how we might interact with it. To challenge assumptions. People like to focus on the "crazy stuff" but Sandy Pentlands work looked barking mad to me in 1997. By 2019 I agreed with him; he basically invented most of modern consumer technology (smartphones, ar, smartwatches, tablets).
