The way I see most new AI projects coming out is they are castles without a moat, what does that mean? Normally when you create a product you want it to last, you want to beat your competitors to the market and you want to have the best product on the market that people are willing to pay for.
If you want to get people to invest in your company then it helps if you have defensible intellectual property (IP) that is either trade secret tech, proprietary processes, or includes patents; without these, anybody can just copy your idea. A lot of the AI use cases that we are seeing come out are built on existing AI models, meaning anybody can replicate it quite easily.
Take GPT Travel Advisor, for example, a great platform for a travel niche. If this had come out a year ago people would be tripping over themselves to have a web app like this. 6–12 months ago a company like Expedia would have paid millions of dollars to acquire something like this to make sure they have an edge over their competitors. Nowadays something like this can be made in an afternoon and finetuned over the following weeks.
As you can see it is a very basic website and in the back end it connects to GPT3 with a prompt like “Write me a travel itinerary for ___ days spent in __________. Break it down into individual days.” and it would generate the itinerary for you.
As you can see the output for Christchurch isn’t exactly the best with a bit of fine-tuning it could be quite awesome. Remember GPT3 doesn’t access the internet yet so they could either wait for the live version or fluff around building a tech stack to fact-check things live, the smart money is waiting for someone else to do the work.
I can see a lot of investors getting into hot water similar to the dot com era when people would invest in “any random idea dot com”. Right now programs like ChatGPT are fun to use but they aren’t the easiest to tailor for individual businesses leading to a wide-open opportunity for newcomers to come and create more tailored solutions.
These tailored solutions will work well in the short term to get people going but as big players bring in entire ecosystems or better integrations into existing workflows the companies that once had a nice little business will quickly fade away. Soon we will see players like Microsoft incorporate a lot of AI into their Office suite which will help.
Long-term ~95% of AI companies are going to rely on the large AI models from companies like OpenAI vs trying to create their own. This means any company can easily copy each other so it could be AI that has a race to the bottom, but in the short term is a race to raise capital and acquire as many users as possible.
Look at the food delivery services, VCs have been subsidising the cost of delivery to scale while trying to capture market share instead of creating a viable and profitable business. The moment they stop the discounts the users switch services or stop using it altogether leading to a collapse.
There is no doubt in my mind we are going to see a massive influx of VCs investing in AI because they see the future use cases but they need to understand that 90% of all of these companies will be wiped out in the long run. A lot of these smaller AI startup companies don’t even need to be acquired by the bigger competition because they have zero IP and unless they have millions of users they don’t hold much value in their business.
