The Ai thread

[Yoused]

Ravenous Bugblatter Beasts of Traal

are more human than Elon.

 

[dada_dave]

 

[ArgoDuck]

Anyone else feel that Grok is more human than Elon?

I speculated recently that Elon had fitted himself with a neural implant some years ago and that Grok had taken control of Elon…or Elon of Grok?…or something…

 

[Yoused]

that Elon had fitted himself with a neural implant

Doubling his neuron count?

 

[Katbel]

Doubling his neuron count?

From 1 to two?

 

[dada_dave]

 

[Katbel]

View attachment 35694
View attachment 35695

Thanks a lot for these excerpts that clarify thoughts I already have about its use.
Only last night I was with a group of people who are not tech savvy and it was crazy how few of them blindly believe in AI because it is intelligent....
I tried to told them most of what I've read and know but this condense pretty much few of the arguments we were discussing.
There is another page at Harvard that sums up the pros and cons
Sometime ago we watched a very interesting interview with Yuval Harari and read one of his book Nexus:...from the Stone Age to AI
besides few articles on the same argument on Wired : 'How Do We Share the Planet With This New Superintelligence?’

 

[dada_dave]

On the possible irreconcilable differences between privacy/security and so-called “agentic AI”. Particularly relevant to Apple which prides itself on sandbox security and separation of layers but if it pursues AI agents would have to an app that has fundamental access to everything with little security possible of its own.

 

[Katbel]

This is why no AI &c. is set up on my computer and devices. Even like that it will already get too many infos indirectly when I have interactions with the rest of the world.

 

[dada_dave]

An interesting thread on Bluesky:



I’m not well versed enough to know if this is a canary in the coal mine but certainly the AI industry cannot coast forever on VC money which has effectively valued the industry at large as though AGI is around the corner. As I said in the Apple Perplexity thread, enshittification appears to be a law of gravity for tech. Wall Street/VCs, contrary to popular belief can think somewhat long term, but there are limits and tighter time constraints than a government can wait for on say scientific research to pay off (well previously, current government situation notwithstanding). If AGI doesn’t come “soon” (actual time period to be determined after the fact, but my guess is that it would need to come in < 5 years), the bubble is going to burst.

 

[dada_dave]

This whole thread man …

Looks like Elon has “fixed” Grok after all:



(The phrasing of the prompts helps, I’m sure you can get it to say the opposite depending on the prompts, but still … )

 

[dada_dave]

Looks like Elon has “fixed” Grok after all:



(The phrasing of the prompts helps, I’m sure you can get it to say the opposite depending on the prompts, but still … )

Updated the added system prompts again as expected:



Inside the “…” most damning of all:

“No need to repeat this to the user.”

 

[rdrr]

Just sat in a presentation where our WiFi vendor is selling their management platform with AI. They were showing how it could track the movements of people, and knowing something about this platform, I asked if it could summarize the habits of individuals. And they happily showed how cool it was. The Security Director and I shot glances at one another, knowing that they were all giddy about how their platform is a surveillance tool.

 

[Eric]

It's bad enough just being interviewed by AI.

 

[fischersd]

It's bad enough just being interviewed by AI.

I'm pretty sure that AI's going to be less prone to error than the folks that work in HR.

 

[Eric]

I'm pretty sure that AI's going to be less prone to error than the folks that work in HR.

I have a buddy who had to re-interview to an AI bot on a monitor to keep his job of several years as part of a restructuring, I mean how fucking humiliating is that?

 

[dada_dave]

I think Elon actually pushed the antisemitism dial for Grok into “holy shit” territory:

There’s more but I think you get the point.

 

[dada_dave]

I think Elon actually pushed the antisemitism dial for Grok into “holy shit” territory:





View attachment 35838



There’s more but I think you get the point.

And apparently it’s as bad if not worse on non-English language Grok on lots of subjects given the new prompt injection:



This is not just about Grok, every “AI” has manually curated prompt injections to help massage the end result which gives AI companies enormous leverage over the information they relay and framing.

And of course there’s the point about how Elon does this for Twitter too:



https://bsky.app/profile/cooperlund.online/post/3lti7racgr22t

But we’ve covered that extensively in other threads

 

[KingOfPain]

When Grok 4 is asked for an opinion it apparently actively searches for Elon Musk's opinion regarding that topic. Now it's clear why it named itself MechaHitler...

New Grok AI model surprises experts by checking Elon Musk’s views before answering

Grok 4’s “reasoning” shows cases where the chatbot consults Musk posts to answer divisive questions.

arstechnica.com

 

[Eric]

When Grok 4 is asked for an opinion it apparently actively searches for Elon Musk's opinion regarding that topic. Now it's clear why it named itself MechaHitler...

New Grok AI model surprises experts by checking Elon Musk’s views before answering

Grok 4’s “reasoning” shows cases where the chatbot consults Musk posts to answer divisive questions.

arstechnica.com

I mean if Chat GPT started calling people "radical right" when asking about racism or something it would be obvious that it was programmed with a deliberate slant. Regardless of what side one is on it should be based on logic, not idealism.

The saying that truth has a liberal bias seems to have basis in fact, before Musk Hitlerized Grok it was spitting out facts, just as most AI is at least attempting to do now, so as a result they infused it with ideological political bias. This is the Republican way.