Given that DeepMind have focused on high predictability machine learning with huge datasets, I wonder if they are best placed to solve healthcare problems. DeepMind algorithms may make accurate predictions, but it will be almost impossible to prove causality, like putting the data into a black box and getting an answer. This is a particularly trixky issue in a high accountability industry like healthcare.
I don’t understand how they can gather enough high quality data to make this anymore than a marketing ploy to differentiate themselves to LPs. Start-ups are by definition new and unproven, data is scarce and often incorrect. A lot of these start-ups are making bets that the future will not be the same as the last, so is historical data really a strong indicator of success?
CrowdFlower is a very interesting company and one that I find hard to predict will succeed or fail. Founder Lukas Biewald is clearly a visionary thinker, having started the company in 2007 when machine learning projects were disregard to research projects that will not be commercializable in the near future.
I believe they have moved away from purely being a crowd-sourced data labeling company and have built out considerable in-house data science capabilities. They claim there million-plus labelers are still hugely valuable, but it will be interesting to see how much they rely on them versus their own people in the future.
I also wonder how the view things such as reinforcement learning and transfer learning as potential disruptors to their business model.
I read recently that quant funds performed very badly during the recent market dip, hinting at them only really being useful in bull markets. I wonder if Quantopian is another example of this. If they are paying out to developers without seeing how these algorithms fair over a sustained period of time, they may be overpaying for the algorithms value.
Great to see another von Ahn post!
I’d be really interested to hear your thoughts on why this ‘twofer’ business model has not been created more often.
P.S. I originally came across von Ahn and his work in the book Big Data (https://www.amazon.com/Big-Data-Revolution-Transform-Think/dp/054422775). If you haven’t read it, I would highly recommend it.
At one stage I thought that GitHub would struggle to get a firm grasp of the enterprise market. We used GitHub in my first job post-University in 2011. My company was a startup, and Github seemed like it was build for startups and individual developers as a collaboration tool. I just did not understand what value enterprises would see in it given that they are often working on highly-proprietary projects and there are a plethora of development collaboration tools built into existing enterprise offerings.
Over time though, I realized that developers have a powerful voice in enterprises. Most developers work on projects outside of their day job and thus most will use GitHub at one time or another. So, when these same developers are advocating for a development platform for their companies, they will push for GitHub.
Interestingly, there are parallels to be drawn with this and Microsoft’s Office strategy in the 1990’s. Microsoft effectively gave Office away to many schools and universities so that their students would become so accustomed to using Office products, that they would only want Office when they entered the work place.
I wonder at what stage Apple will have to sacrifice some of the ‘quality’ for quantity. Its US iPhone sales have begone to flat-line, and they are unlikely to see much growth in the Western World (in terms of ecosystem membership, not handset refresh). So as they aggressively target Eastern and Developing economies, will the have a suite of apps that meets the needs of these very different consumers?
I wonder how susceptible Stack Overflow would be to disintermediation in its enterprise play. Companies who receive high value answers/information from certain contributors, may want to use those contributors as consultants or hire them as employees.
I think it is all but impossible for Dropbox to move to Blockchain, as its current technology is pretty much the antithesis. Dropbox has made its money on an entirely centralized file storage concept. Although it will have distributed storage farms for performance and redundancy, in essence it centralizes all users data within its black box. Dropbox offers no customization or visibility.
If Dropbox moved to Blockchain, it would have to do so gradually, as all current user data could not be redistributed amongst users who have not opted in to sharing their computing resources. This dual approach could not work against companies that are all-in of P2P file storage.
Alternatively, if Dropbox just shut down its cloud service and started from scratch, it would have huge fixed assets that would be sold at a margin of their price and nothing more than its brand as a ‘cloud’ file storage provider. This would leave way for Google/Amazon/Microsoft to pick up Dropbox’s crumbs.
For Dropbox to bet on block-chain would be for it to bet against itself. I agree with Haley; this is a classic Innovator’s Dilemma and one that I don’t see an answer to.
Wonderfully entertaining topic right now Julia.
Google’s TPUs are arguably more effective for Deep Learning purposes but it has yet to be made available outside of the Google Cloud Platform, will be interesting to see if Google try to compete here.
Also, Intel’s response to nVidia has been to partner with AMD (historically it’s biggest rival) to develop GPUs that can compete.
It shall be an entertaining battle.
Your eloquent assessment of Netflix’s success was a thoroughly enjoyable read; a feat befitting an individual as refined as you.
I was particularly surprised that over 50% of Netflix’s content is now original. We have learned in DI&T in recent weeks that it is often beneficial to move from a product to a platform, opening up to an ecosystem while making that ecosystem reliant upon you. It appears that Netflix is doing the opposite. It does not have a technology advantage, so it is creating original content and no longer wants to be the medium for disseminating content from all other content creators.
I agree and I offer you the sincerest apologies if that came across. I do not believe on-prem is dead, but I believe on-prem alone is dead. Hybrid models will continue to exist as highly regulated industries, especially in the EU, will need to store data locally; but there will be very few (if any) large enterprises that will not be cloud-first.
The problem with Oracle is that its legacy sales culture does not work; a hybrid model cannot have a fixed, 7 year contract without modularity or dynamic computing resources. Microsoft still have a huge on-prem offering, but that is mainly sold through resellers as opposed to a direct sales force, leaving Microsoft more able to dictate its new sales culture.