AI: Greatest Apocalypse of our times?

[IndianRenegade]

So this happened.

 

 

[ravishingravi]

 

[ravishingravi]

Jurassic park scene captures effectively what we are dealing with here. 

[Tillu]

On 9/9/2026 at 12:24 AM, IndianRenegade said:

If AI can do this, on it own, something the brightest of humanity haven't for a hundred of years. What chance do much of humanity has? 

 

Reports are now coming out that Ai was trained on so many mathematicians previous works and unpublished works.

At this moment Ai cannot independently solve an entirely new problem on its own unless it was trained on it.

 

It was a PR job by openAi to give everyone an impression that their Ai could solve even the most difficult mathematical problems on its own. It's a good productive assistant for now which can quickly sieve through large sets of data built upon someone else's work.

[IndianRenegade]

4 hours ago, Tillu said:

 

Reports are now coming out that Ai was trained on so many mathematicians previous works and unpublished works.

At this moment Ai cannot independently solve an entirely new problem on its own unless it was trained on it.

 

It was a PR job by openAi to give everyone an impression that their Ai could solve even the most difficult mathematical problems on its own. It's a good productive assistant for now which can quickly sieve through large sets of data built upon someone else's work.

This will need to play itself out. But let's assume the worst-case scenario, OpenAI trained on unpublished mathematicians work, brute forced it to a solution. 

 

Most knowledge work today are orders of magnitute simpler than this, there is a vast corpus of data available on them that humans voluntarily gave up helping the opensource community, now that data will be used to put the ppl who created the data out of their livelihood. The coming years are going to be brutal unless the society understands what they have unleashed & works on a path forward.

 

 

Edited Thursday at 06:49 PM by IndianRenegade

[Tillu]

5 hours ago, IndianRenegade said:

This will need to play itself out. But let's assume the worst-case scenario, OpenAI trained on unpublished mathematicians work, brute forced it to a solution. 

 

Most knowledge work today are orders of magnitute simpler than this, there is a vast corpus of data available on them that humans voluntarily gave up helping the opensource community, now that data will be used to put the ppl who created the data out of their livelihood. The coming years are going to be brutal unless the society understands what they have unleashed & works on a path forward.

 

 

 

It didn't solve the problem on it's own. Two mathematicians used the help of Ai tools to arrive at this solution. It required human input to direct it towards the solution. openAi spied on these people's work and breakthroughs, plagiarized it and passed it off as their own achievement.

Edited Thursday at 11:52 PM by Tillu

[randomGuy]

I think if today or anyday in future, AI wants to k!ll us, it will. This tech has happened in last 2-3 years. It's not in hands of humans now. But we should still try by pausing and keeping praying.

 

 

 

 

Edited Sunday at 06:12 PM by randomGuy

[Tillu]

20 hours ago, randomGuy said:

I think if today or anyday in future, AI wants to k!ll us, it will. This tech has happened in last 2-3 years. It's not in hands of humans now. But we should still try by pausing and keeping praying.

 

 

 

 

 

It's so uncertain whether to believe the current Ai Hype vs Reality. While it certainly made some tasks easier it still has a long way to go.

 

For instance, if you take even Software Engineering for example, it solved only a small part of the entire Software Development Life Cycle.

 

Out of the many phases in a typical SDLC, it solved only the coding part which is not really a big thing actually because even before AI, people were copy pasting code from Google and StackOverFlow.

 

Now whatever productivity gains we got from generating code is being spent on the tiresome process of reviewing it. So coding was never a bottleneck.

 

End of the day, these Ai tools like Claude, ChatGpt etc are probabilistic models and not deterministic. It still needs human supervision to verify and review the results.

Edited Monday at 02:19 PM by Tillu

[randomGuy]

25 minutes ago, Tillu said:

 

It's so uncertain whether to believe the current Ai Hype vs Reality. While it certainly made some tasks easier it still has a long way to go.

 

For instance, if you take even Software Engineering for example, it solved only a small part of the entire Software Development Life Cycle.

 

Out of the many phases in a typical SDLC, it solved only the coding part which is not really a big thing actually because even before AI, people were copy pasting code from Google and StackOverFlow.

 

Now whatever productivity gains we got from generating code is being spent on the tiresome process of reviewing it. So coding was never a bottleneck.

 

End of the day, these Ai tools like Claude, ChatGpt etc are probabilistic models and not deterministic. It still needs human supervision to verify and review the results.

 

My brother is a tech lead involved with a leading US bank. He says almost all the coding is through AI (they have claude subscription and use up all tokens by the end of the month and would be happy to be provided more tokens). Bug fixes ,unit tests etc all being done with AI. He says work has become a lot easier and manpower requirements is less and has been cautioning me against buying IT shares from past 1 year. You are right, my brother says often it hallucinates, provides wrong fixes, needs human supervision.

 

We as humans will k!ll someone to protect ourselves from danger. It depends if AI is conscious and self protective. Humans may only know the answer after it's too late. I think we are already too late. We need good luck and as well as pausing immediately. I don't think china will play ball. It's a difficult situation. Far more risky than nukes. Have been saying this from past year.

[kepler37b]

Google gemini is frustratingly inaccurate for market research.

 

It reports public companies as private, fails to get the latest movement in asset prices and thinks that Decathlon sells home cleaning appliances. 

 

It is good for general idea, please check and re-check again for final reports. 

 

It does the basic stuff very well, but when it comes to things that we can rely on, it is hot or miss.

[kepler37b]

4 minutes ago, randomGuy said:

 

My brother is a tech lead involved with a leading US bank. He says almost all the coding is through AI (they have claude subscription and use up all tokens by the end of the month and would be happy to be provided more tokens). Bug fixes ,unit tests etc all being done with AI. He says work has become a lot easier and manpower requirements is less and has been cautioning me against buying IT shares from past 1 year. You are right, my brother says often it hallucinates, provides wrong fixes, needs human supervision.

 

We as humans will k!ll someone to protect ourselves from danger. It depends if AI is conscious and self protective. Humans may only know the answer after it's too late. I think we are already too late. We need good luck and as well as pausing immediately. I don't think china will play ball. It's a difficult situation. Far more risky than nukes. Have been saying this from past year.

It is a blasphemy to doubt on AI these days. That is corporate hysteria.

 

Let the dust settle down. 

[Tillu]

10 hours ago, randomGuy said:

 

My brother is a tech lead involved with a leading US bank. He says almost all the coding is through AI (they have claude subscription and use up all tokens by the end of the month and would be happy to be provided more tokens). Bug fixes ,unit tests etc all being done with AI. He says work has become a lot easier and manpower requirements is less and has been cautioning me against buying IT shares from past 1 year. You are right, my brother says often it hallucinates, provides wrong fixes, needs human supervision.

 

We as humans will k!ll someone to protect ourselves from danger. It depends if AI is conscious and self protective. Humans may only know the answer after it's too late. I think we are already too late. We need good luck and as well as pausing immediately. I don't think china will play ball. It's a difficult situation. Far more risky than nukes. Have been saying this from past year.

 

Yeah it can generate code for everything, but what if Ai is leaving backdoors within the code. So we now have the added overhead of reviewing the code.

 

And generating code alone is not sufficient for low latency applications. So we also have to benchmark the code against different API to achieve low latency.

 

If we give complete autonomy to Ai all the institutional knowledge will be lost and it becomes highly impossible to debug and maintain 100s of 1000s of lines of code.

 

Even Indian IT companies have become clever in how they bill their clients. Now instead of billing per hour they are billing per every task completed.

 

Yes Ai deflation is a thing but not to the extent it will replace every IT job. Having good domain expertise will easily land you jobs even today.

[ravishingravi]

Very base fundamental question beyond the risk

 

 

[Tillu]

1 hour ago, ravishingravi said:

Very base fundamental question beyond the risk

 

 

 

In a perfectly utopian society people will start living a life of leisure but geopolitics is a zero sum game, whoever controls the chips, data centres, fabs, lithography machines and that whole ecosystem will control the world.

 

ASML a company in Netherlands which builds the Lithography machines,

 

Zeiss a German company which supplies mirrors, a key component in these machines,

 

TSMC which manufacturers 90% of the semiconductor chips in the world,

 

Nvidia which designs the GPUs,

 

Korean companies which manufacture the memory chips

 

and the Japanese who provide the chemicals for doping the chips

 

 are the key players in this ecosystem and US controls every one of them. China is the only country looking to build all this IP on its own without much success.

Edited 22 hours ago by Tillu

[BacktoCricaddict]

4 hours ago, ravishingravi said:

Very base fundamental question beyond the risk

 

 

 

I think about this 24/7. One of the best classes I taught - and I did this from 2013 - 2016 (much before AI) - was in a flipped format."

 

I had my tech serv folks video-record all my lectures before the semester and posted them on our learning management system. Before each class period, students would watch one of these 80-min lectures and answer a set of questions pertaining to it (I called them objectives). During our class meetings (and attendance was optional), we would do a Q/A, which would then lead to exploration of the topics at higher levels. We were not wasting time in class on content delivery, rather using class time for higher-level synthesis and analysis. 

 

After the first exam results, attendance was never a problem. 

 

Funding ran out to update the video recordings ... but with AI, I am thinking of reinstating this method. 

[IndianRenegade]

Brutal Oracle Lay offs!

[Lord]