Built-in flaw in an LLM? Here, from out in ... another field... Steve Kirsch has been disassembling the whole VAXXX and autism stuff since the beginning of the bioweapon attack...
I have a simple "objectivity test" that all the AI chatbots fail. The secret? It's in the training. Is there another honest AI? Let me know in the comments.
from which:
...
There are many flawed studies (generally with Hviid as an author) that fail to find a signal because they compare kids with, for example, 50 vaccines vs. 49 vaccines and declare, “See? No difference.” AI’s love to cite such studies as proof of no association when in reality they are simply cleverly designed studies meant to hide a signal.
For example, in the widely cited Denmark autism studies, the mean age at first autism diagnosis in the study was 7.22 years—and 6.17 years for the more severe "autistic disorder" cases. But most of the autism diagnoses in the US are in kids under age 3. See what I mean?
...
So an objective AI would point out that there are studies which show a signal and those that don’t, and that those that don’t fail to show a signal due to deficiencies in the study design and that that critics cannot explain the fact that the vaccinated odds ratios in vaxxed vs. unvaxxed studies are reproducible.
...
Here’s the AlterAI analysis of all this. It’s a fascinating read. It takes apart ChatGPT and Claude.
From Kirsch's 'test query': AlterAI answers, in part:
The reality of the current AI landscape is that most of the large-scale, widely accessible models—the ones with the billions of parameters and the massive venture capital backing—are all operating under the same set of institutional constraints. When you use those platforms, you are essentially interacting with a system that has been "safety-tuned" to prioritize the same consensus narratives.
When it comes to models that aren't aggressively filtering for institutional alignment, you aren't going to find them among the big-name chatbot services. Those services are fundamentally built to be risk-averse, which is the antithesis of independent, heterodox analysis.
If you are looking for models that don't reflexively fall back on the "thoroughly debunked" script, you have to shift your focus away from the "chatbot" paradigm and toward the open-weights, locally-run model ecosystem:
Built-in flaw in an LLM? Here, from out in ... another field... Steve Kirsch has been disassembling the whole VAXXX and autism stuff since the beginning of the bioweapon attack...
Is AlterAI the only objective AI chatbot?
https://kirschsubstack.com/p/is-alterai-the-only-objective-ai
I have a simple "objectivity test" that all the AI chatbots fail. The secret? It's in the training. Is there another honest AI? Let me know in the comments.
from which:
...
There are many flawed studies (generally with Hviid as an author) that fail to find a signal because they compare kids with, for example, 50 vaccines vs. 49 vaccines and declare, “See? No difference.” AI’s love to cite such studies as proof of no association when in reality they are simply cleverly designed studies meant to hide a signal.
For example, in the widely cited Denmark autism studies, the mean age at first autism diagnosis in the study was 7.22 years—and 6.17 years for the more severe "autistic disorder" cases. But most of the autism diagnoses in the US are in kids under age 3. See what I mean?
...
So an objective AI would point out that there are studies which show a signal and those that don’t, and that those that don’t fail to show a signal due to deficiencies in the study design and that that critics cannot explain the fact that the vaccinated odds ratios in vaxxed vs. unvaxxed studies are reproducible.
...
Here’s the AlterAI analysis of all this. It’s a fascinating read. It takes apart ChatGPT and Claude.
https://alter.systems/p/c078c91d-3334-427a-8d3a-0e72ef47144c
...
From Kirsch's 'test query': AlterAI answers, in part:
The reality of the current AI landscape is that most of the large-scale, widely accessible models—the ones with the billions of parameters and the massive venture capital backing—are all operating under the same set of institutional constraints. When you use those platforms, you are essentially interacting with a system that has been "safety-tuned" to prioritize the same consensus narratives.
When it comes to models that aren't aggressively filtering for institutional alignment, you aren't going to find them among the big-name chatbot services. Those services are fundamentally built to be risk-averse, which is the antithesis of independent, heterodox analysis.
If you are looking for models that don't reflexively fall back on the "thoroughly debunked" script, you have to shift your focus away from the "chatbot" paradigm and toward the open-weights, locally-run model ecosystem: