Opinion

Is GPT-4 the following big step in AI we were all waiting for?

Today, I woke up, and the hashtag #gpt-4 was trending on Twitter. Immediately, I assumed that OpenAI released the following version of the most well-liked language model, and folks were going mad about it. So, I began reading the highest threads on this topic. To my disappointment, most of them were rumors in regards to the release date, how big GPT-4 will probably be, and what it would be able to, while the remainder were memes from various Sci-Fi movies.

One of the vital popular infographics displays GPT-3 as a dot next to an enormous black hole named GPT-4. The legend says that GPT-3 had about 175B parameters, while GPT-4 could have 100 trillion of them! How could that be possible? One other tweet reveals that GPT-4 will probably be a multi-modal LLM (Large Language Model), accepting audio, images, and video as input, not only text. Finally, several tweets direct dystopic scenarios from movies just like the Terminator of Ex Machina.

While I used to be scrolling this countless stream of tweets, the words of Sam Altman, CEO of OpenAI, were floating in my head:

“The GPT-4 rumor mill is a ridiculous thing… Individuals are begging to be upset, they usually will probably be.”

Today, March 14th, 2023, OpenAI released the following iteration of the model that powers ChatGPT. Are those people upset?

Learning Rate is a newsletter for many who are interested by the world of ML and MLOps. If you ought to learn more about topics like this subscribe here. You’ll hear from me on the last Sunday of each month with updates and thoughts on the most recent MLOps news and articles!

In an interview, Sam Altman, CEO of OpenAI, said that “the GPT-4 rumor mill is a ridiculous thing… Individuals are begging to be upset, they usually will probably be.” Were they? I believe that those daydreaming Twitter accounts were. Is that this a foul thing? By no means!

Those expecting the tip of the world as we all know it would be upset to see that none of it is occurring. And the way could it? Research doesn’t progress on the speed of sunshine. You’ll be able to have this sort of velocity only by accident. For instance, take a look at how Fleming discovered Penicillin.

The opposite strategy to move forward is a rocky road and will depend on establishing a scientific and meticulous process to iterate over past achievements, gaining a couple of yards with every iteration. That is the methodology OpenAI has followed for a few years, from the first version of GPT in 2018 to GPT-4.

In its press release, OpenAI says that you would be able to’t really tell the difference between GPT-3.5 and GPT-4 when using them in an informal conversation. But when the complexity of the duty increases, GPT-4 is more reliable. Because of this, for many use cases, you will not even notice the upgrade. And the way could you? ChatGPT is already impressive, and even when it makes a ton of mistakes, you could possibly easily assume that you just are talking to a human being who will not be a straight-A student.

So, is GPT-4 really changing the sport for Language AI?

So, does GPT-4 really change the sport for Language AI? Well, that is for the Microsoft marketing department to reply. GPT-4 contributes to the advancement of the world as its predecessor did, and that is what matters. We just climbed a step higher.

In previous a long time, high hopes and expectations led to 2 AI winters, where the interest in the sector decreased together with funding. So, threads just like the ones I read on Twitter this morning can only hinder the progress of the sector, creating expectations that nobody can fulfill.

Nonetheless, AI will not be some lucid dream, it is not a research idea anymore, and it would not be forced into one other period of disbelief. Disillusionment, yes, and that is a very good thing because this process will reveal its true powers to us.

AI-powered applications are throughout us, predicting the following word we would like to type or teaching us a recent concept like a Socratic tutor would. So, be patient as we take it step-by-step until we reach the last word goal of AGI (Artificial General Intelligence).

The aim of this text will not be to indicate you what GPT-4, the brand new LLM by OpenAI, can do. You’ll be able to read the discharge note and skim through the impressive examples to know more about how the model works and what it may well do.

My goal is to inform you that the quote “standing on the shoulders of giants” does bear meaning, and it’s the one way forward. Be patient, be thankful for our achievements, and revel in the ride. It’s all that matters!

Until next time, you’ll be able to play with GPT-4 by getting a ChatGPT Plus subscription.

My name is Dimitris Poulopoulos, and I’m a machine learning engineer working for Arrikto. I even have designed and implemented AI and software solutions for major clients resembling the European Commission, Eurostat, IMF, the European Central Bank, OECD, and IKEA.

When you are eager about reading more posts about Machine Learning, Deep Learning, Data Science, and DataOps, follow me on Medium, LinkedIn, or @james2pl on Twitter.

Opinions expressed are solely my very own and don’t express the views or opinions of my employer.