The Good and Bad of Elon Musk’s Grok AI

To say Elon Musk is an interesting guy would be an understatement. He goes from brilliant to bozo and back again in a blink. It’s like watching a fast ping-pong match.

Musk’s latest effort is an AI based on X (formerly known as Twitter) content called Grok that I believe will disrupt conversational AI.

Shaking up a new market is a good thing in many cases because, too often, everyone starts copying each other, and you end up with a lot of similar and increasingly boring offerings. While Musk does a lot of things, he doesn’t do boring, except, of course, the company by that same name.

Let’s chat about Grok this week. Then, we’ll close with my Product of the Week, the Surface Laptop Studio 2. Microsoft sent me one, so now I have experience using it.

The Good

We don’t seem to like to make fun of things, which I think is a shame. While I don’t like many things about Tesla, I do like that it has a touch of whimsy in its execution that must bring a smile to the faces of Tesla drivers. For instance:

• The Tesla can make Jetson car sounds.
• You can change the car on the navigation screen to a variety of crazy vehicles.
• It can make a fart sound when someone sits in one of the seats.
• You can do karaoke in the car — Tesla calls it Caraoke.
• Tesla has a “bioweapon defense mode,” which is really just a good in-car air filter, but I love the name — and a ton of Easter eggs, like James Bond’s submersible Lotus.
• Tesla was the first to allow you to use your phone or a card as the car key so the doors open when you approach.
• It has Dog Mode to keep your pets cool in the car on a hot day.
• You can block explicit song lyrics if kids are in the car — or leave them unblocked to teach your neighbor’s kids new words!

So, there is a really good chance Musk’s Grok AI will be fun to use. Assuming it eventually works as advertised, since most products like this have early teething problems and crashes, I expect this could be a great way to lighten your mood while traveling or killing time.

We need more whimsy in our lives since so much of the news is negative. Two major wars are going on, both of which could escalate and spread to nuclear weapons, and politics have gotten so ugly it is almost painful to read or watch the news now. So, most people should appreciate a little humor in this Debbie Downer world (SNL reference).

However, Grok also has some problems.

The Bad (There Is No Ugly)

First, you must be an X Premium member to use Grok. That’ll cost you $16 a month. ChatGPT costs $20 a month, but it appears to have far more substance behind it, and it’s flowing into Microsoft Office via another monthly extra charge.

Grok’s announcement indicated the AI will include real-time knowledge from X/Twitter. However, that information rarely contains the answers to questions (with respect to the overall traffic). Since Musk largely cut out moderators on the platform, its accuracy and quality have degraded while the individual protections from being improperly attacked have evaporated. That means the X training data is likely horribly compromised.

Grok may be funny, but if you want it to do actual work, it may not be very helpful and could get you in a lot of trouble if it pulls from inappropriate Tweets from a hostile foreign national or someone who doesn’t filter what they post.

Interestingly, Musk was one of the people who thought we should do an AI pause because he thought the technology was very dangerous. While this is a concern I share, an AI trained from Tweets could be the very danger he is afraid of. It’s as if he wants to create a self-fulfilling prophecy by creating an AI that could do much harm if trained on a massive amount of hostile, questionable human and bot actions and behaviors.

You want an AI to be accurate and honest, and I don’t think that’ll be possible with one trained on Tweets because Tweets are often inaccurate or dishonest. If people make decisions based on this or this data migrates to other training sets, we could have a lot of whacked-out AIs. I doubt that’ll be good for any of us.

Wrapping Up: Grok, for Better and for Worse

Part of me really wants to play with Elon Musk’s Grok, but I worry about what it will do and how what it does will reflect on me. For instance, were I to ask it a legitimate question, how do I know it won’t pull content from those who believe in Q and that anyone seeing my screen will not think I’m some kind of whack job?

Still, I really like the idea of making one of these things fun to use, but much like the folks who allegedly played with real guns on that film site, this could become far more dangerous over time and put us at higher risk.

People say some pretty nasty things about politicians and the inappropriate use of weapons and explosives, and there is a wide variety of hate speech on the platform, any of which showing up on your business laptop could get you fired in a hot moment and, depending on where you work, even land you in jail.