Software is in fact "eating the world". Software as a Service (SaaS), Cloud Computing, AI/ML, Productivity Software, countless embedded processors with software running all sorts of machines, are examples of software technologies that have, and will continue to, impact our businesses and personal lives.
In a recent episode of the All In Podcast the team talks about the emergence of new Artificial Intelligence (AI) technologies like ChatGPT, and the potential implications on the global search business. Its an interesting discussion that offers insights on the search business, its underlying technologies and the potential future for AI models like ChatGPT.
Notably, Microsoft recently made a follow on investment of $10 billion into OpenAI, creator of ChatGPT, in exchange for a significant ownership percentage of the company and exclusive rights to the ChatGPT technology.
At first, it would seem the availability of a superior solution for language processing would be an immediately disruptive force in the search market, perhaps giving Microsoft, a traditionally distant competitor to Google, a leg up and an opportunity to quickly capture market share.
But a more detailed discussion of the search business, and some key metrics involved quickly reveals that it might not be that simple.
In the podcast, David Friedberg (min 37:00), a former Google insider, walks through some of these metrics based on a back of the envelope analysis.
Using public revenue numbers, he estimates that Google generates ~$2 to 3 of revenue per Ad click on the company's platform. With a click through rate of roughly 3% for presented ads (per search), he arrives at an estimate of ~$0.05 to $0.10 of revenue per search. Working with the $0.05 estimate, and assuming a 50% cost (COGS) per search, the cost to Google per executed user search comes out to about $0.02 to $0.03.
So what goes into the cost of executing a single online search? Along with operating expenses like software development, sales and other administrative functions, a key cost driver is the hardware processors (depreciated capital expenses charged to COGS) tasked with executing the data analyses, along with the electricity needed to power them.
The challenge to ChatGPT, he asserts, is that based on recent studies, the computational requirements for a comparable search using ChatGPT are nearly 10x more than what is required using current search technologies, or about $0.30 per executed search.
Of course, costs will come down as learning models improve, and as computing power and energy progress down their own development cost curves.
But it is a good reminder that every piece of software runs on a piece of hardware, and how software performance and capabilities are inextricably linked to to the cost, performance and capabilities of the underlying silicon hardware.
And for ChatGPT, not to worry, as The Innovator's Dilemma would instruct us, this is how many disruptive technologies get their start.
Commercial Fusion Research Company Spins Out Power Innovations for e-Mobility, Storage
Commercial Fusion innovator TAE Technologies has announced they will be spinning out subsidiary TAE Power Solutions to commercialize its new battery technologies for the e-mobility and energy storage markets.
The new technology was develop as a direct result of the company's fusion research when the local Southern California utility was unable to meet the peak power needs of their latest research reactor.
The answer was a highly scalable battery technology capable of storing and quickly discharging large amounts of power.
The full article is available here:
New Funding and Other Updates
