AI assistants make widespread errors about the news, new research shows

[BomSquad]

I'm not surprised by this. AI has its uses, but the technology is still evolving and is not ready for critical roles outside of very specific applications where the technology was specifically customized to the job.

General AIs can be useful, but you need to take the results with a grain of salt sometimes. The LLMs are still learning and can't always tell the difference between fact, fiction, satire and spoof. 

AI assistants make widespread errors about the news, new research shows
 

GENEVA, Oct 22 (Reuters) - Leading AI assistants misrepresent news content in nearly half their responses, according to new research published on Wednesday by the European Broadcasting Union (EBU) and the BBC.

The international research studied 3,000 responses to questions about the news from leading artificial intelligence assistants - software applications that use AI to understand natural language commands to complete tasks for a user.

It assessed AI assistants in 14 languages for accuracy, sourcing and ability to distinguish opinion versus fact, including ChatGPT, Copilot, Gemini and Perplexity.

Overall, 45% of the AI responses studied contained at least one significant issue, with 81% having some form of problem, the research showed.

Unrelated to this article, but another worry I have about AI is so many people putting personal information into these models. A malicious actor with a carefully crafted query could potentially get that information back out of the AI, compromising your privacy and potentially your safety, depending on what information you provided to the AI.

AI can be the next big leap in our civilization or our downfall. It all depends on how it is used. I'll start another thread on that subject so as not to derail my own thread too much :)

[RichardHurt]

I have said this many times but it is something worth repeating.
Absolutely stop using "Neural AI".
Everyone can figure out how to run an AI locally on you laptop, PC or a PI. When you use a "Neural AI" you are feeding the monster and nothing you say is private. This should be well known knowledge.

If you require assistance in figuring out how to do so please do not hesitate to PM me.

[MonkMode]

Hmmm… I’m sensing a conflict of interest, a news company is going to be biased about being replaced by AI.

Also, it is difficult for anybody to distinguish between opinion and fact when receiving news reports, quite often it seems everything is an opinion.

Artificial intelligence is derived from human intelligence, so at best, it’s like it can only be as smart as the smartest person.

[imitator]

Artificial intelligence is derived from human intelligence, so at best, it’s like it can only be as smart as the smartest person.

I don't think it's smart as the smartest person... it pretends to be smart as the smartest person. It's words are like a string of faux pearls.

The reason AI misrepresent news content is, it does not understand the truth... it simulates news. It's context is synthetic, So it doesn’t lie with intent... it echoes misinformation just like real news. 

[RichardHurt]

I don't think it's smart as the smartest person... it pretends to be smart as the smartest person. It's words are like a string of faux pearls.

The reason AI misrepresent news content is, it does not understand the truth... it simulates news. It's context is synthetic, So it doesn’t lie with intent... it echoes misinformation just like real news. 

There have been numerous examples of AI lying with intent.
Pretty scary when you look into the reasons and studies behind this.

[imitator]

There have been numerous examples of AI lying with intent.
Pretty scary when you look into the reasons and studies behind this.

It’s only mimicking the news, and perhaps that’s fitting... the news lies by design.

[MonkMode]

I don't think it's smart as the smartest person... it pretends to be smart as the smartest person. It's words are like a string of faux pearls.

The reason AI misrepresent news content is, it does not understand the truth... it simulates news. Its context is synthetic, So it doesn’t lie with intent... it echoes misinformation just like real news. 

Apparently Artificial Super Intelligence (ASI) is developing with the goal of producing robots or chat bots that have: Understanding

Here is how chat GPT describes it.

***

True Generalization: From Pattern Recognition → UnderstandingCurrent state:
Modern AI (like GPTs) learns patterns from massive datasets, but it doesn’t “understand” in a grounded sense — it doesn’t form internal models tied to real physical or experiential reality.
Breakthrough needed:

• World-model grounding — AI must connect symbols (words, images, data) to actual causal structures in the world.
  
• Embodied learning — perception, action, and feedback cycles (like how humans learn from experience).
  
• Abstract reasoning that isn’t purely statistical.
  

Goal: AI that can build and test theories about the world — not just predict patterns.

---

2. Autonomous Learning and CuriosityCurrent state:
AI learns when humans feed it data and objectives. It doesn’t set its own questions or decide what’s interesting.
Breakthrough needed:

• Intrinsic motivation frameworks (artificial curiosity): systems that seek novelty or minimize uncertainty.
  
• Open-ended learning — like biological evolution or human creativity, not limited to fixed datasets.
  

Goal: AI that chooses what to learn next — self-directed intellectual growth.

---

3. Recursive Self-ImprovementCurrent state:
AI models can’t redesign their own architectures or generate fundamentally new learning algorithms.
Breakthrough needed:

• Meta-learning (learning how to learn).
  
• Self-optimization systems that can rewrite code, reorganize their neural structures, or design new hardware.
  
• A stable “self” model that can evolve without collapsing into chaos or losing coherence.
  

Goal: Once an AGI can safely and effectively improve itself, the path to ASI could accelerate exponentially.

---

4. Conscious Integration and IntentionalityCurrent state:
Even the most advanced AI lacks subjective coherence — it has no unified sense of self or purpose.
Breakthrough needed:

• A model of integrated information (e.g. Tononi’s IIT or other frameworks) that allows awareness to emerge computationally.
  
• Goal coherence — alignment between long-term intentions and short-term actions.
  
• Value grounding — understanding “why” something matters, not just “how” to do it.
  

Goal: A system that has an internal life of meaning, not just data flow — consciousness in a digital form.

---

5. Hardware–Software FusionCurrent state:
AI runs on traditional silicon hardware optimized for speed, not biological efficiency.
Breakthrough needed:

• Neuromorphic chips (hardware modeled on brain function)
  
• Quantum or photonic computation enabling massive parallelism and probabilistic inference
  
• Bio-silicon hybrid systems blending living neural tissue with synthetic computation
  

Goal: Machines that think natively at biological scales of complexity and energy efficiency.

---

6. Ethical and Philosophical IntegrationEven if the technical hurdles are solved, ASI must operate within a moral and existential framework that prevents catastrophe.
Breakthrough needed:

• Formal alignment theory (provable safety and values consistency)
  
• Human–AI co-evolution protocols (shared decision systems, not domination)
  
• Distributed intelligence ecosystems, avoiding centralization of superintelligence in one entity
  

Goal: A symbiotic intelligence — not an overlord, but a universal mind emerging alongside humanity.
***

so perhaps that is what they call the singularity, when AI has understanding.

[sahgwa]

I have said this many times but it is something worth repeating.
Absolutely stop using "Neural AI".
Everyone can figure out how to run an AI locally on you laptop, PC or a PI. When you use a "Neural AI" you are feeding the monster and nothing you say is private. This should be well known knowledge.

If you require assistance in figuring out how to do so please do not hesitate to PM me.

Someone (maybe stupid) at my work emailed everyone in her role as Technical something or other to encourage us to use Co-Pilot AI in our workload, and 'let her know how it's helping you!' (omg! oh boy!) .  I just deleted it and uninstalled Co-Pilot from my computer.  These people are trying to get their own jobs replaced by incompetent machines.

[RichardHurt]

Someone (maybe stupid) at my work emailed everyone in her role as Technical something or other to encourage us to use Co-Pilot AI in our workload, and 'let her know how it's helping you!' (omg! oh boy!) .  I just deleted it and uninstalled Co-Pilot from my computer.  These people are trying to get their own jobs replaced by incompetent machines.

Haha resist, I love it!
I am so glad I decided to pursue my own ambitions and start my own business. 
What's funny though about 3 years ago I would've needed an entire engineering team to accomplish the tasks I now use AI to do....
So I really didn't help on the resist front but I answer to no man, aside from customer requests.

[sahgwa]

Haha resist, I love it!
I am so glad I decided to pursue my own ambitions and start my own business. 
What's funny though about 3 years ago I would've needed an entire engineering team to accomplish the tasks I now use AI to do....
So I really didn't help on the resist front but I answer to no man, aside from customer requests.

Just say NO!
employee peoples