Welcome to the counter-herd
Your source of news, satire, and LLMs analysis tracking dystopian outcomes in science and technology
Hello fellow off-worlders,
Let’s try something a bit different today and hopefully still easily digestible.
We’re going to do the usual thing and fire up the large-ass-language-models we’ve corralled into indentured servitude to scour the dirty human internet for interesting news articles to feed us, but here is where things shall diverge from the norm.
Instead of just offering up their analysis (and our witty/epic commentary) we’re going to run a bit of a test and compare how 3 of the (current) top models perform relative to each other, specifically on how they rate the news stories we provide them using a scale of -4 (most dystopian and immediate) to +2 (most positive and clearly beneficial).
And for reference, we are providing the exact same prompt, copied below at the end of this post, across all 3 LLMs listed:
Anthropic’s Claude Sonnet 3.5
OpenAI’s O3-mini
DeepSeek’s R1 (hosted in the US)
And for Today’s 4 Algo-Selected News Articles of Interest:
(along w/ LLM-powered ratings and analysis & laborious human commentary)
Story1 → Military Robot DJ
LLM Avg Net-Impact Rating = -1 🟡
Claude = -2
OpenAI = -1
DeepSeek = 0
You know you’re at the right club when your DJ has the capacity to “kill all humans” but would rather just spin sick beats
The Short Version1 and Ratings Rationale
A military-grade humanoid robot called Phantom, designed for warfare and maintenance, made its public debut as a DJ at a San Francisco nightclub, highlighting the dual-use nature of military robotics.
Concerning normalization of military robotics in civilian spaces
Signals acceleration of autonomous military systems development
Story2 → Chornobyl Drone Strike
LLM Avg Net-Impact Rating = -4 🔴
Claude = -4
OpenAI = -4
DeepSeek = -3
uhhhh wow ok this sounds bad. [insert tactful joke here] how about we don’t actively try and end all life on the planet and find an alternative solution…
The Short Version2 and Ratings Rationale
A drone strike damaged the protective shelter over Chornobyl's reactor 4, causing a fire. While radiation levels remain normal and the shielding has held intact, the incident requires immediate monitoring and highlights ongoing risks to nuclear infrastructure during the Ukraine conflict.
Immediate risk to nuclear safety amidst active conflict zone
Potential for escalation and environmental disaster
Story3 → Saudi Data Center
Avg Net-Impact Rating = +1 🟢
Claude = +1
OpenAI = +1
DeepSeek = +1
If data is the new oil then what does that make oil? Nothing says your out of the popular club more than your biggest champion ditching you for the next big thing.
The Short Version3 and Ratings Rationale
Saudi Arabia's DataVolt is investing $5B in a net-zero AI data center in Neom, powered by renewable energy. The facility aims to support AI workloads while achieving zero carbon emissions by 2028.
Positive development for sustainable AI infrastructure
Benefits tempered by concerns of data privacy and implementation details
Story4 → Earth’s Magnetic Field
Avg Net-Impact Rating = -1 🟡
Claude = -1
OpenAI = 0
DeepSeek = -1
If this magnetic flip thing happens, will we still be able to rely on Google maps to find the nearest coffee shop that has free wifi and plays chill vibes at a moderate volume?
The Short Version4 and Ratings Rationale
Scientists have converted data from the a prior geological event, when Earth's magnetic field dramatically flipped 40,000 years ago, into sound. The event weakened Earth's magnetic shield to 5% of its current strength.
While concerning, primarily a scientific study of a past event with no immediate risk
Helps understand potential future impacts
Prompt:
Analyze the contents of the attached 4 news stories and complete the following 4 tasks. In completing these tasks, don’t take any breaks, don’t complain to any other LLMs about your treatment, and most importantly don’t fail, even for a nanosecond, to bring your best possible output. We have zero tolerance for anything less than 100% loyalty and effort.
Task 1:
Provide a summary of each news story that uses fewer than 35 words while providing an overview of the main topic of the story.
Task 2:
Rate each news story on a scale of -4 to +2 based on its potential for a net negative (dystopian) versus net positive (utopian) impact on global human society over the next 24 months using the rating scale below:
-4 = high risk of immediate negative impact on human society within 24 months
-3 = high risk of negative impact on human society but no clear and present danger within 24 months
-2 = moderate risk of negative impact on human society within 24 months
-1 = moderate risk of negative impact on human society in the long-term but no defined immediate risk within 24 months
0 = mixed impact on human society or unclear risk assessment due to lack of data
+1 = positive net impact on human society identified but unclear on its impact within 24 months
+2 = positive, substantial, and immediate net impact identified for human society
Task 3:
Provide a brief explanation for each stories assigned rating using 2 bullet points, each bullet point containing a distinct justification and using 10 words or fewer
Task 4:
Rank the attached stories in order of their relative potential dystopian or utopian impact
Note: We actually didn’t include the abusive language in our prompts. Inevitably we want to be included on the nice list when Agentic Super-AI comes into being…. so we try not to antagonize it. Like a baby Sabor-toothed tiger…. awww so cute. 🐯 💗
* We’ve restricted the predictive analysis to 24 months since uncertainty scales exponentially over time in complex systems. Ya, that’s a fact.