Grok 3 - does it deliver? 👀

Happy Tuesday Non-Techies,

The beta version of Grok 3 was released last week. I mentioned Grok a few times in last week’s chatbot crash course, but in case you’ve already forgotten (no judgement here), Grok is xAI’s (uncensored) horse in the chatbot race.

Is it the scruffy, stubborn mule that likes to play dirty? Or a polished stallion that’s already galloping ahead?*

In today’s newsletter, I’ll try to answer that question. First, a quick poll:

Okay, let’s get started. For those about to Grok, I salute you (sorry, I couldn’t resist).

What makes Grok 3 stand out?

Until now, Grok’s rebellious sass was probably its biggest differentiator. If we ranked chatbots based on how likely they are to get caught smoking in the school playground, Grok would win hands down. A solid 9/10 on the snark scale.

That ‘personality’ is still there with Grok 3, but according to xAI, it has other unique selling points too:

🧠 It’s now really smart. 
🖥 It’s fuelled by a giant supercomputer called Colossus.
🤔 It uses a step-by-step reasoning process, kinda like a person might.
🎓 It can act as an academic researcher too.
🆇 It pulls real-time information from the web and from X.

Let’s dig further…

___

🧠 How smart is Grok?

If only there was a score to measure this…oh wait, there is! The Chatbot Arena might sound like some sort of 2025 Robot Wars reboot, but disappointingly, it’s just a platform that tells you which chatbot is the smartest.

It does that with a score called an ELO. Users blind-vote on their favourite answer produced by anonymous chatbots and points are awarded to the winner.

No chatbot has ever achieved an ELO score of over 1400 before…until now.

But it can’t take all the credit. You know what they say: behind every great chatbot is an even greater computer. Let’s meet Colossus. 

🖥 Colossus is…colossal.

With a name like Colossus, you’ve really got to deliver the goods. Well, Colossus does. It’s believed to be the world’s largest AI supercomputer, and boy is it chunky. 

Here’s a glamour pic:

Researching this thing is a minefield for non-techies, but here’s the big takeaway: it means Grok 3 is 10x more powerful than Grok 2, capable of solving more complex and intricate problems.

It’s kind of weird to remember that this technology has this physical presence. Lest we forget the environmental impact this all has too. 

___

🤔 Reasoning like a person.

OpenAI’s o1 and o3 models and Deepseek pipped xAI to the post on this one, but it’s undeniable that Grok 3 likes to show off. Where other chatbots and earlier versions of ChatGPT will spit out an answer, Grok 3’s Think mode shows you its workings along the way.

🎓 Academic-level research.

The same is true for its DeepSearch mode, which is its research function (apparently as clever as a PhD level researcher). Here, rather than just a thought process, you’ll see a sequence of search queries, sources and conclusions. Grok 3 wants you to know where it gets its information and the logic it’s using to give you an answer. As it says itself, “This isn’t about getting to an answer, it’s about transparency.” Fancy pants.

Be warned - it isn’t quick - these new research functions can take minutes to get a good answer for you, not seconds. 

___

🆇 X marks the spot.

Apparently (and when I say ‘apparently’, I mean ‘according to Google’s AI summary’), around 500,000,000 posts are published to X every day. It won’t come as a shock to you that X’s chatbot offering has been given a front-row seat to this trove of breaking news, opinion, sentiment and…let’s face it, quite a lot of drivel.

The supposed benefits are access to real-time information, social insights, a more diversified pool of knowledge and the ability to cross reference its web sources with X posts. Quite how useful this is, we’re yet to see. 

So, is Grok 3 my new squeeze?

Honestly? Maybe I’m uptight (or maybe I’m just not a beer-chugging frat boy) but I’m not really drawn to chatbot that can yell and call me a tw*t and engage in sexy voice messages. That stuff puts me off. 

That’s not to say it doesn’t have serious power - as it now clearly does. It’s easy to use and it’s just as strong, if not stronger, than most of the stuff out there. 

It’s almost like the more techie you are, the more excited you are about Grok 3. Developers, computer scientists, mathematicians…these demographics seem to be the most impressed by Grok 3’s extra computational power (courtesy of our old pal Colossus).

But for me, as a non-techie, 26 more ELO points than ChatGPT and an extra serving of sass probably isn’t enough to turn me into a Grok 3 fan girl yet. 

I’ll keep an eye on it though, for sure.

Okay, hopefully you know a bit more about Grok 3 now. Let’s do another poll:

Right then, it’s Sunday as I write this newsletter, so I’m off to make an “epic breakfast” for my extended family - I’m talking pancakes, sausages, bacon, pastries, fruit, all the good stuff. 

Have a great week!

Heather

*bonus pic of the scruffy mule vs the sleek stallion. Bit creepy? Reply and let me know.

PS I’ll see some of you in today’s 3 day Become an AI Trainer bootcamp, starts at 5pm GMT (tickets still left if you want to join last minute)