Catching the ChatGPT Wave
Large Language Models and Industry Research
The last time I was this excited about a new technology was 1992 when I discovered the internet. I was reading an interview by William Gates III (not to be confused with Bill Gates who would discover the internet three years later). This Gates had a popular print magazine called Midnight Engineer which occasionally published my first essays. He interviewed a startup entrepreneur and asked him “What has had the biggest impact on your success?” I remember my reaction to the answer: “The Internet.” I wondered “What’s the Internet?” (Back then Internet was capitalized.)
Within a week I had signed up for Delphi Internet, a dialup service that connected to a VAX VMS system that gave you access to Archie, Veronica and WAIS. By the time I launched an ISP, RustNet, in November of 1992, the first version of Netscape had been released by a kid out of Illinois named Marc Andreessen. The rest is history.
This time around I don’t have to dump a career to jump on a speeding freight train. Last March IT-Harvest had launched a platform for researching 3,000+ cybersecurity vendors. How can OpenAI’s GPT3, a large language model (LLM), be leveraged to enhance our offering?
It took four weeks for Maximillian to launch v4.0 SOCrates. The first task was to write 3,200 summaries of what each vendor did. Here is an example:
Fortinet is an American multinational corporation headquartered in Sunnyvale, California. The company develops and sells cybersecurity solutions, such as physical firewalls, antivirus software, intrusion prevention systems, and endpoint security components.
Sure, I could produce a better description. But it would take 1,500 hours to write 3,200 such snippets. It took GPT3 a total of 150 hours (6+ days) to do the work. But that is not all it did.
GPT3 provided all those links to Wikipedia, FaceBook, even Github. And it collected the locations of all of Fortinet’s major offices. It would take our team of 11 in India over a week to do that for all the vendors with manual searches and cutting and pasting into a spreadsheet.
Continue scrolling on a vendor page in the platform and you see:
That is the normal view of headcount over time and breakdown by country where Fortinet employees live. But the new data appears if you open the drop down for Leadership. The amount of data collected is so extensive I don’t want to publish it publicly.
Now look at the investment-by-round data that we track. This is for 1Password.
I knew Ryan Reynolds was an investor because of the commercial he recorded for 1Password. All the investors are scrollable in the platform. And yes, you can pivot on any investor to see their entire portfolio of cybersecurity investments.
That is what Ashton Kutcher’s portfolio looks like.
Here is how a big investor’s portfolio appears.
But wait, none of these are what drove us to latch on to the power of GPT3. The game changer is that we have ingested products and descriptions for all the vendors.
Not perfect, but still pretty amazing. (Note the “Beta” designation.)
We have the data, now what about searching? Click here for an interactive walk through of the product search feature added five days ago.
There are 690 vendors and 827 products with the terms “AI” and “ML” in their descriptions. There are NONE that mention ChatGPT. That will change.
By now you are thinking “How do I get access to all this data?” Simple. Just visit dashboard.it-harvest.com and sign up for a monthly subscription. For other plans reach out and we can talk.
Richard Stiennon