Altman acknowledged this week that his company is bumping up against some limits. While underlying AI models are “still getting better at a rapid rate”, he told reporters at a San Francisco dinner, chatbots like ChatGPT are “not going to get much better”.
There are lots of good comments here, and as the writer, I do appreciate all of them.
One element of this story, which was hard to really spend the time on, is the public's difference in perception regarding disruption of other technology companies, compared to its perception of industrial and service based industries. The public doesn't seem to care when a company like Intel takes on a company like Fairchild Semiconductor. Part of the issue is a lack of technical sophistication, so it is difficult to separate competitors based on their products. The more important reason, though, is that technology devouring technology is easily understood as progress. 500 engineers lost their job at one firm, but a new firm is hiring 500.
Now take a look at the service disrupters like AirBnB, Uber, etc. First, unlike technical disruption, the public understands the businesses here very well. It's a hotel. It's a taxi. It's a laundromat. Second, there is a distinct feel that these new companies are not playing by the rules, whatever those rules might be (it doesn't help that these companies publicly flaunt the rules either). Third, and most importantly, there is far more perception of the people losing their jobs, rather than the gain these companies are making in terms of labor flexibility.
There are plenty of greenfield companies out there (Nest, DeepMind, Climate Corporation are acquisitions in the last month that come to mind). But the region is not exclusively doing that kind of progress anymore, and so we shouldn't be surprised when people aren't immediately positive about the changes taking place anymore.
The book, "Anatomy of an Epidemic", provides a really comprehensive look at how we ended up in this situation. Some of the keys here:
1) A desire to bring a level of "science" to a part of our physiology we don't understand. The thinking here is that while we do not understand the etiology of depression, we can at minimum begin to use blunt tools to solve problems. The issue as anyone who has studied complex systems understands, is that the feedback loops are so dense, there is no method to understand what is happening.
2) Financialization of treatment – drugs make more money than therapy and other methods. Or to use an HN phrase, drugs are more easily scaled to the population than other methods. The incentives throughout the entire system push people this direction, regardless of the underlying research.
3) Treatment doesn't happen instantly in any case. The issue with much of the research today is that we take a very limited time window to evaluate the efficacy of different treatments. If, instead, we looked at treatment over the life course, the results are often radically different.
This is where startups like Seven Cups of Tea will hopefully play the world. This mental health crisis offers a huge opportunity for disruption and creativity. As a quote in the Stanford alumni magazine said this month: "One hundred years from now, people will look back at the age of giving SSRIs and they will have a reputation that's akin to bloodletting."[1]
One other consideration is the number of active accounts in various countries. I imagine the penetration in the United States is one of the highest, and the US population is fairly large as well. It would be interesting to get some sort of a relative factor here to get a sense of how aggressive these governments are seeking information.
As a developer on the Python stack, I would love to know when would be a good time to start using Go in serious production work. It seems to me that it solves a lot of the backend services infrastructure problems associated with interpretive languages (one of the reasons I was considering diving in Scala or other JVM languages), is relatively reliable, and has a fairly strong core library. It still seems bleeding edge, but the language seems to have developed far faster than Python did over the last decade or so.
When considering this type of transparency, there is a spectrum of types of people. Some prefer radical transparency where every possible pitfall and challenge is openly acknowledged and noted. On the other side of the spectrum, a lot of people don't want to have any knowledge of the challenges, and in fact, work most productively in a state of naiveté.
Part of hiring is understanding how different people respond to this sort of stress. For me, I really prefer this sort of off-the-record honesty. I know a start-up is challenging, and a failure to acknowledge a reality that I know exists is a huge turnoff. But I can definitely imagine that some hires would find this scary and would find a startup that seems much more "perfect." You attract who you want to attract.
Got to hang out with the folks at Rough Draft last night – I am glad to see that their passion for helping entrepreneurs is finally coming to fruition. Keeping in mind that these are students making these investments (many who were simultaneously studying for midterms this week), it is truly inspiring to see such energy in the Boston ecosystem. Keep up the good work!
Thrun seems to be getting into this mode as well with this line: "He’s thinking big now. He imagines that in 10 years, job applicants will tout their Udacity degrees. In 50 years, he says, there will be only 10 institutions in the world delivering higher education and Udacity has a shot at being one of them. Thrun just has to plot the right course."
Why is it that people (even apparently faculty like Thrun) seem to forget that universities deliver more than just undergraduate education? "Delivering higher education" also means research labs and centers for scholars, graduate education, professional education, executive education, etc. While undergraduate education may be ripe for disruption, there is a serious leap needed to go from there to the complete disappearance of thousands of institutions.
Just look at the numbers: dederal research grants and endowments will sustain at least several hundred universities for the long-term, and other universities that are currently teaching-focused will move more of their efforts to research as students take online classes and stop being paid customers.
The institutions that should be worried are technical colleges (depending on major), community colleges and perhaps state college systems. They are the most likely to be disrupted, particularly if Udacity could offer a more comprehensive curriculum. But Stanford? Or even schools like the University of Florida or UC-Santa Barbara? They have plenty of other income sources, and still have a lot of life in them.