Interior of Big Modern Server Room. Illustration 3D
aging
Right now, many people are obsessed with the diagnosis of AI’s condition. How far have we come to exploring the big linguistic models?
We know that things change quickly. People are catching on to the numbers that show not only the way AI makes in a market and business framework, but how experts predict their growth in the next two quarters until we hit 2026.
First, there is this set of numbers that I heard come out of a recent AI Daily Ship episode, where Nathaniel Whittemore collected at least 15 diagrams to pass the nuts and bolts of AI’s adoption. You may be forgiven that you are thinking that a podcast is a strange environment for charts, since your audience can really not see what you are looking at, but I just wanted to present some of the numbers as an example:
- Ghatgpt goes to 100 million users in just a few months, earning previous platform files like Tiktok
- The rapid escalation of ANTHROPIC revenue
- Google’s doubling in contract processing within two months
- Increasing companies moving from experimentation to AI development from 11% to 33% in two quarters
- 45% of a group of workers who have been investigated who is willing to interact with AI agents at work
This is a general overview of what is happening in the world AI. Clearly, we are in a time of great change.
Notes from IEEE
The Institute of Electric and Electronic Engineering is a big name in technological analysis. With about 460,000 members in 190 countries, the commercial team pays attention to those trying to read tea leaves in technological development. I looked at a article by Eliza Strickland providing more data points to AI. One thing that IEEE’s resource entered was the cost of falling the conclusions, where the number was cut in half in a few months.
Among other findings: that Chinese models attract American, that companies are being made on board with AI lining and that LLMs are becoming more and more experienced in clinical diagnosis.
The last one?
Another item covered by the IEEE piece was the performance of AI to something called The last examination of humanity, that sounds ominous in a clock. HLE is a deep crowdsourcing comparative assessment material, from 1,000 expert partners and 500 institutions in 50 countries. See how the project is inserted on its own website:
“Reference points are important tools to monitor long -language developments (LLM). Designed to be the final academic reference point of the closed type with wide coverage.
So, basically, AI’s leaders are trying to create stronger, more difficult tasks of comparative evaluation, because AI is doing so well with existing ones.
What IEEE found, in Hle’s point of view, is that even top models like O1’s O1 are still in the single digits when they are about to get the right answers, which experts point out is a good sign for the utility of the challenge.
Is Grok ahead?
I also wanted to include one more data set, This time from AI seasonwhere presenters ranked individual models according to their skills at Teraflops. Someone stood out: Take a look at the Grok 3 model, which is removed to Scattershot.
And then think about this: Since July 9, Grok 4 is already out.
In July 10 coverage in Ai GyaniShikha Singare tells us how Grok sits AB the Leaderboard Arc-AGI-2.
“Grok 4 … (delivers) ~ 18% accuracy, clearly in front of Claude Opus 4 (~ 10%) and GPT-4.5 or O3 variants,” Singare writes. “While Grok 4 is more expensive per job, leads to raw performance-an urgent choice when the accuracy of the work is critical to the shipment.”
Singare also brings something called Vending Bench, a dynamic market simulation, where the model did extremely well.
“Grok 4 achieved a stunning average net value of $ 4694.15, far ahead of Claude Opus 4 ($ 2077.41) and even human participants ($ 844.05),” Singare adds. “He also sold the highest number of units on average: 4569, indicating strong generation of demand.”
Were some dozen ideas? Not everyone was in sphere points. I think if you put all this together and connect the dots, you get an exciting picture of how AI is going to accelerate the second half of this year and to the next.
