Aurora Magazine

Promoting excellence in advertising

AI, the Meta Way

What is Meta AI and how can we use it?
Published 23 Apr, 2024 12:02pm

Meta, the company that owns Facebook, Instagram and WhatsApp just launched their AI model, Llama 3, through their apps, and the whole nation has gone berserk using its playful generative AI feature, “imagine.”

Basically, you can now go to any of the Meta apps and simply type something like, “Imagine Pakistani women at the International Space Station,” and it will instantly create an image based on your prompt. This is what I got:

For some reason, Meta’s AI thinks a lot like Pakistani men because even at the International Space Station (ISS), the AI image generator has Pakistani women making gol (round) chapatis!

But besides that minor technical glitch bordering on ultra-conservative male chauvinistic stereotyping, you have to be amazed at how the learning model behind the AI engine works. So many questions rush to my mind!

How did the AI know what Pakistani women look like? How did it understand what the ISS is? How was it able to put this realistic image together in less than five seconds? How did it know that Pakistani women should be covering their heads with dupattas when in a non-mehram outer space environment? Is Llama 3 Shariah-compliant? How did it know of the central role a round chapatti plays in a typical Pakistani woman’s life, even at the ISS?

You will also be happy to learn that Meta AI is not just restricted to generating Shariah-compliant images of futuristic Pakistani women. As we speak, the AI can help you with any of six tasks:

1. Ideas: For example, it can help generate ideas on how to put together the perfect beach party for your client.

2. Writing: Help make your emails more professional or write the copy for your next ad.

3. Fun: Play games like trivia or tic-tac-toe.

4. Support: Help you cheat on your college assignment (we are not advocating this) or build your resume (we are advocating this based on the quality of CVs we received from recent Pakistani business school graduates).

5. Learn: Help simulate a job interview or even a rishta interview with your future father-in-law!

6. Imagine: Generate images based on prompts like we did in the example above. Hours of fun are guaranteed.

To the average social media user, this update might seem like nothing more than another gimmick, and to some, even an irritation. But behind the scenes, there is a full-fledged AI war shaping up between tech giants. A war that will shape the future of humanity.

If you think that is an overly dramatic assessment, consider that the top leaders who worked for decades to create artificial intelligence models suddenly panicked one day and started pushing the US Congress to pass restrictive regulations on AI to “protect humanity.”

Prophecies of entire career paths becoming redundant, making large swathes of the population unemployed are actually the least scary of these predictions. If AI is integrated into any system that controls even the simplest of weapons, for example, imagine the outcome of a rogue, self-aware or even defective AI entity.

There is a reason why I am discussing these points. Unlike Google and OpenAI, who are now closely guarding their code because of these threats, Meta decided to release their code to the public, making it truly open source. Some of you may argue that this is what ChatGPT also did. In fact, ChatGPT’s parent company, OpenAI, put the word “open” in its name because that was their vision: to make the code available to humanity.

But something went wrong that created a whole lot of panic for Sam Altman, CEO OpenAI. Altman has now restricted access to GPT 4’s code, the latest version of ChatGPT, and is advocating for laws and regulations to control AI. If you are a follower of AI news, you will have heard that Elon Musk is now suing OpenAI for violating its vision by restricting access.

But why is Meta making its AI model, Llama 3, open to the public? As is the case with anything that Zuckerberg and Co. does, the reasons are commercial. By making their code available to the public, they are betting that their model will become the preferred base for developers to build apps and chatbots on. But is this greed? Is Meta compromising safety over money-making?

Ahmad Al-Dahle, VP of generative AI at Meta, addressed this concern in the last Llama release. “We take these concerns seriously and have put a number of things in place to support a responsible approach to building with Llama 2.”

Yet, shortly after Meta released its first Llama model to the public on February 24, 2023, the code leaked on the open forum 4Chan, where it was used by bad actors to spread misinformation and build chatbots that propagated offensive material, including racial hatred and violent graphic scenes.

But not everyone thinks Meta is wrong with its open-source approach. Nazneen Rajani, research lead at Hugging Face, another open-source AI platform, is excited. According to Nazneen, “We will be able to uncover more secret ingredients about what it actually takes to build a model like GPT-4.”

Whatever the case may be, AI is no longer a thing of the future; it is the present and has already started impacting all of us. The difference that Meta AI will make, however, is to help fuel the growth of AI by making it accessible to everyone.

AI before this moment was, through models like ChatGPT, visible to a certain group of people, such as students or professionals like advertising creatives. With the Meta AI update now, every Tik-Tok-Bukhari-Commercial-vlogging-Pakistani-influencer type with a WhatsApp account has been given access to artificial intelligence. This will either turn out to be a very good thing, making Pakistan great again (MPGA), or it could just as easily be the start of an apocalypse. Time will tell.

Syed Amir Haleem is CEO, Skale Interactive and Content Fusion.