Why ChatGPT Search is Better than Google
I don't know how ChatGPT Search works, but having built a similar search engine myself, I feel qualified to have assumptions about it. Therefore, to explain how ChatGPT Search works, I must use our own system as an example - Assuming that the developers at OpenAI built something similar, since it's simply the superior way to build an AI-based search engine. Look carefully at the following screenshot.
First of all, it's performing multiple searches to create a comparative analysis. In each search it will autonomously decide which keywords to use. Each search invocation will return up to 20 links, and we allow the AI to autonomously decide which of these links to scrape to answer the original question. This process allows the AI to prioritise and read the articles it believe is most important to do its job. The LLM will then aggregate the most relevant articles, allowing it to answer the user's original question. If we imagine a human being doing the same job, we'd end up with something resembling the following:
- Construct multiple different search queries likely to give you development speed facts about each programming language.
- Use Google to search for results using the above queries.
- Prioritise the search engine's result to find the articles that are most relevant to answer the original question.
- Read these articles.
- If the human still require more information it would have to perform additional searches until it's found the information required to answer the original question.
- Answer the user's original question by writing a summary.
Doing the above manually would easily require 5 minutes to construct the optimal search engine queries, 5 minutes to perform the searches, 5 additional minutes to find the most relevant links to read, and some 25 minutes to read all the relevant articles. Then add some 20 minutes to write a summary based upon the information gathered, and you've now spent 60 minutes doing the same job manually.
What takes you 60 minutes with Google can be done with ChatGPT Search in 60 seconds!
Google Cannot Create Working Software
Google has 20,000 software developers and engineers, all of whom have been thinking about nothing but this problem for 2 years. Still, one single software developer (me) was able to implement it almost a year ago, and Google has yet to release anything that actually works in this domain. Basically, one single software developer was able to implement something Google couldn't do, outperform Google on their own playing field, and deliver something that's orders of magnitudes better than Google's own products. Basically ...
One software developer outperformed 20,000 Google engineers!
This dilemma isn't unique to AI search, it's everywhere in Google. For instance, Google's reCAPTCHA library is basically junkware. If you include reCAPTCHA on your page, you've increased bandwidth usage by 1.7MB. I implemented a PoW-based CAPTCHA library using 5KB. In addition you can see Google's lack of quality in their products. For instance, I don't understand Greek, still about 50% of my current YouTube ads are Greek. Google is supposed to know everything about me. Still, for "mysterious reasons", they cannot even serve me ads in a language I understand.
Why Google is failing and what we can learn from it
At some point Google stopped caring about code quality, because they were too busy thinking about "everything else". I'm not going to go into a political discussion with my readers, but there are important facts we can learn from Google's failure. To summarise what we can learn, please realise the following;
Google is failing because delivering high quality products for the least amount of resources is no longer their priority.
The software development industry in general is experiencing similar problems, because we've become obsessed with process and axioms, having lost focus on delivering high quality products. In the following video I explain why Google is failing, and how it is relevant to your company, what we can learn from it - In addition to how our ChatGPT Search module works.
The Whatever Frikkin' Works Software Development Process
Over the last couple of months, I have created a video series explaining what's wrong with the software development industry. In the series I go through all of our "best practices", explaining how they're counter productive to our ability to delivering working software, with my intentions being to have you see the world from a more pragmatic point of view. My process implies the following changes:
- Drop unit testing for your indeterministic code.
- Stop using Git for all your code. There exists code where using Git is not useful.
- Drop your CI/CD pipelines and save straight to production when you can.
- Forget OOP, it's anyways nothing but a software development mass psychosis.
- SOLID, Clean Architecture, and Design Patterns are useless.
- DDD is software development schizophrenia.
- Event sourcing only adds overhead and makes your project more costly without adding any real value.
- Etc, etc, etc ...
The series is work in progress, and I still have more subjects to cover, such as NoSQL and Event Sourcing - But feel free to watch the series below if you're interested in how you can improve your product quality while simultaneously reducing resource costs to deliver products.
Have a Custom AI Solution
At AINIRO we specialise in delivering custom AI solutions and AI chatbots. If you want to talk to us about how we can help you implement your next custom AI solution, you can reach out to us below.