In an AI-Content Dominated World, What Happens to AI Itself?
Building Startups by Ajay Yadav
The ‘No-BS’ BS Newsletter
Happy Friday once again👏
Let’s go🚀
💻️AI
➡️ In an AI-Content Dominated World, What Happens to AI Itself?
Since the dawn of the Internet, it is humans who have produced the majority of content available online.
Be it articles, blogs, product descriptions, ads, even e-mails— while a tech tool may have helped us groom and refine them, they were at the end of the day human products.👨
Now we know Generative AI models have been trained on huge datasets (mostly human-generated) scraped from the Internet.🌐
But the early-adoption of Gen AI models has led to a significant amount of AI-generated data to be present online, and is serving as a major source for training these models. 🖥️
So, naturally, an intriguing question arises:
How will GPT change as LLMs contribute most of the language we see online? ⌨️
A group of researchers explored exactly this. In a research paper titled The Curse of Recursion: Training on Generated Data Makes Models Forget, the authors say this “loop" leads to a phenomenon called Model Collapse. 🤖
Model Collapse is a degenerative process in which models forget the true underlying data distribution over time, even when there is no shift in the distribution over time.
It begins with the tails of the distribution disappearing, and over time, the learned behaviours become more and more similar, converging to a single point with very little variation.🔻
So how do we avoid model collapse?
By having and continuing to provide LLMs access to real human-generated data.🙇♂️
Simply put, if LLMs are used extensively to generate content on the Internet, it could contaminate the dataset used to train them.
So to ensure that learning continues for an extended duration, it's essential to maintain access to the original data source, the human data, and also ensure that supplementary data not created by LLMs remains accessible over time.⌛️
➡️ AI News of the Week: DALL-E 3 is Now Available to ChatGPT Plus and Enterprise Users
Remember when we discussed OpenAI’s very significant announcement of DALL-E 3? Well, DALL-E 3 is now finally made available to all ChatGPT Plus and Enterprise users. 🖼️
All you have to do is share your vision, and ChatGPT will generate many variants for you to review in real time! Read more here.
➡️ AI Tool of the Week: GPT-trainer
I realised I haven’t talked about a custom chatbot builder tool in any of my previous issues, and GPT-trainer helps you do just that in minutes!⏱️
Simply import your data, personalize the chatbot, and put it on your website - it's that simple. You can create a chatbot for any function: HR, business consulting, lead generation, content creation, teaching assistant, language translator, and lots more! 🤖
🚀Entrepreneurship
Lots of people believe that the best way to run a business is the bootstrapped way.
I was asked about my thoughts on this on a podcast recently, and my philosophy is simple, because my previous business taught me this the hard way:
The best way to run a business is to have a business that makes money.💰️
It doesn’t matter if you raise money, it doesn’t matter if you run it bootstrapped. They both have their pros and cons.
With bootstrapped business, you keep all the equity, you keep all the upside. But it might take you longer to reach your goal. ⌛️
With venture capital, you keep some of the equity, and some is with the investors. But, you might just grow faster. ⏩️
Regardless of this, the rule has to be that either of these businesses should be making money, because most businesses never make money 💹
So first and foremost figure out how your business will make money, and then take your call to decide between bootstrapping or raising funds.💲
👔 Leadership Lifestyle
Arya Taware founded her fintech startup FutureBricks straight out of University College London at age of 22 in 2015. The idea was to provide funding for small and medium-sized homebuilders in the U.K. and serve as a portal for property experts and investors to connect.🏘️
She has successfully raised £4 million in equity, built a solid team, board and infrastructure among other things, all while being a first-generation migrant and a single female founder.♀️
Arya’s company has total loan origination of circa £12 million across various projects with a 100% track record of interest and capital payback to date. She is also a feature of the Forbes 30-Under-30 under Finance, Europe.🎖️
She is a true role model for women around the world, and a dear friend. This is what she says about building teams, and something I greatly resonate with:
Have a great weekend folks, see you in the next one!
As usual, share this newsletter with friends through the link below :)