🤖 WTF is AI?

Shedding light on what AI is and what it is capable of and more precisely exploring the potential of the ChatGPT engine.

💤 The buzzword

I know I am late to the party. But there have been profound changes around me in the recent time and I couldn't really concentrate on blogging.

Anyway, I was hoping for Blockchain to be the next blast in the tech industry and in fact, I wrote an entire blog on tech that I believe in for the next decade. But AI was out of the syllabus 😬

So what exactly is AI and why it has created a buzz in recent days?

Many define AI as something that reduces manpower and increases productivity. Of course, AI does that. But that's not how AI can be defined.

AI is the replication of human intelligence (aka human brain) with the help of high-processing computers. This seems to be kind of a gimmick right? But AI is evolving faster than we think. Check out how AI is evolving in this video here.

According to a Bloomberg article, there have been 1,072 mentions of AI on S&P 500 company earnings calls so far this year. According to the National Bureau of Economic Research, Companies embracing generative AI have been earning shareholders 0.4% more per day than those with less exposure to the tech.

⌨️ The next word predictor - ChatGPT

So how can we create a human brain? We need to train the systems with tons of tons of data and these systems try to make a pattern out of it and that's how intelligence is born cause we as humans tend to find patterns in everything we do and that's how our intelligence is defined.

To explain it further, I am going to dig deeper into ChatGPT.

GPT is a Generative Pre-trained Transformer. Transformer is basically you feed in a lot of data as input. The data includes everything right from Food to Quantum Physics. For instance, ChatGPT is apparently trained on Reddit data and Reddit is actually suing Sam Altman (CEO of ChatGPT) for using their data. Okay coming back to the definition of the term - so GPT can produce anything - it can output images, songs, videos. But in the context of ChatGPT - it is capable of building a conversion like a chat in the form of text only.

ChatGPT is trained on billions of words on the internet and when generating answers for the questions that you ask, it tries to predict what the next word is and it ends up mimicking human writing.

So ChatGPT is basically a next-word predictor. Seems very usual right, like a word generator?

But hold on, think deeply - what is next word prediction? If I can predict what next word you are gonna say, then I have gained your intelligence and that’s where I become powerful.

But when I say next-word prediction, the question of the discussion is - Whether the latency is under control. Otherwise, the next-word prediction is totally obsolete, right?

And, I have to give credit to OpenAI for cracking this.

Well, there are 2 types of Network latency (the time a web page was waiting for the connection to be established ) - Actual latency and Perceived latency. Actual latency is the actual amount of time to make an API call to the database and fetch the information. Whereas, perceived latency can be improved a lot in the UI. Because when you prompt on ChatGPT, you already see the information start to stream and you almost felt like the response time was negligible, whereas Google’s Bard couldn’t get this correct as it had to wait for the entire information to be loaded to be displayed, and this gave a perception that Bard was way much slower than ChatGPT. But in the backend, both these LLMs were taking the same amount of response time to be loaded. Tackling this latency is so important for businesses when they are especially focusing on conversion rate metrics.

🙋🏽‍♂️ Million Dollar question

Now shifting gears, the million-dollar question - If we can create a human brain with tons of data, can AI replace humans?

I want to take a pause here and kidnap you to all of the industrial revolutions that happened and you see, in each of the revolutions, there were some innovations that happened that replaced humans. For instance, cars were assembled entirely by humans a century ago, then machines and robots were introduced and now barely a human touches the parts of a car.

Likewise, I don't think that AI will take human jobs and I strongly believe that "History will try to repeat itself".

Take a look at the below data where a normal person would prefer Humans over AI in ~70% of the cases, especially in the fields of Doctors, Lawyers and Financial Advisor where a small error in AI would lead to disaster and a normal person wouldn't love to take that risk. So while AI continues advancing, many remain partial to human judgment in situations where lives, freedom, or finances are on the line.

But at the same time, we definitely need to upskill ourselves cause we need to be efficient enough that we can't be replaced.

I know I am trying to paint a positive picture of AI, but if you wanna see the other side of AI, I highly recommend checking out Varun Mayya's YT channel where he explains why AI is dangerous. One of the most amazing channels that I have discovered on AI.

I have lots to tell like how AI along with blockchain can help gain trust in businesses. In fact, I have a whole theory on that and if that works it's gonna be huge 🤯

But I want to keep this short and stay tuned for much more blogs on AI and tech, and the different AI tools that I have tried out.