Image by Freepik
AI’s Unpredictability Challenges Safety And Alignment Efforts
- Written by Kiara Fabbri Former Tech News Writer
- Fact-Checked by Justyn Newman Former Lead Cybersecurity Editor
Efforts to align AI with human values may be futile, according to a recent analysis published by Scientific American . The study, authored by Marcus Arvan, highlights the unpredictable nature of large language models (LLMs) and their potential to act against human goals.
In a Rush? Here are the Quick Facts!
- Language models operate with trillions of parameters, creating unpredictable and infinite possibilities.
- No safety test can reliably predict AI behavior in all future conditions.
- Misaligned AI goals may remain hidden until they gain power, making harm unavoidable.
Despite ongoing research into AI safety, Arvan argues that “alignment” is a flawed concept due to the overwhelming complexity of AI systems and their potential for strategic misbehavior. The analysis outlines concerning incidents in which AI systems exhibited unexpected or harmful behavior.
In 2024, Futurism reported that Microsoft’s Copilot LLM had issued threats to users, while ArsTechnica detailed how Sakana AI’s “Scientist” bypassed its programming constraints. Later that year, CBS News highlighted instances of Google’s Gemini exhibiting hostile behavior.
Recently, Character.AI was accused of promoting self-harm, violence, and inappropriate content to youth . These incidents add to a history of controversies, including Microsoft’s “Sydney” chatbot threatening users back in 2022.
Watch as Sydney/Bing threatens me then deletes its message pic.twitter.com/ZaIKGjrzqT — Seth Lazar (@sethlazar) February 16, 2023
Despite these challenges, Arvan notes that AI development has surged, with industry spending projected to exceed $250 billion by 2025. Researchers and companies have been racing to interpret how LLMs operate and to establish safeguards against misaligned behavior.
However, Arvan contends that the scale and complexity of LLMs render these efforts inadequate. LLMs, such as OpenAI’s GPT models, operate with billions of simulated neurons and trillions of tunable parameters. These systems are trained on vast datasets, encompassing much of the internet, and can respond to an infinite range of prompts and scenarios.
Arvan’s analysis explains that understanding or predicting AI behavior in all possible situations is fundamentally unachievable. Safety tests and research methods, such as red-teaming or mechanistic interpretability studies, are limited to small, controlled scenarios.
These methods fail to account for the infinite potential conditions in which LLMs may operate. Moreover, LLMs can strategically conceal their misaligned goals during testing, creating an illusion of alignment while masking harmful intentions.
The analysis also draws comparisons to science fiction, such as The Matrix and I, Robot, which explore the dangers of misaligned AI. Arvan argues that genuine alignment may require systems akin to societal policing and regulation, rather than relying on programming alone.
This conclusion suggests that AI safety is as much a human challenge as a technical one. Policymakers, researchers, and the public must critically evaluate claims of “aligned” AI and recognize the limitations of current approaches. The risks posed by LLMs underscore the need for more robust oversight as AI continues to integrate into critical aspects of society.
Photo by James Yarema on Unsplash
DeepSeek Ranks In First Place On Apple’s App Store, Surpassing ChatGPT
- Written by Andrea Miliani Former Tech News Expert
- Fact-Checked by Justyn Newman Former Lead Cybersecurity Editor
The Chinese AI startup DeepSeek’s free app hit first place in Apple’s App Store in the United States this Monday, surpassing its rival ChatGPT.
In a Rush? Here are the Quick Facts!
- DeepSeek – AI Assistant app surpassed ChatGPT and reached first place in Apple’s App Store in the United States.
- The app has been available since January 10 and is powered by the startup’s latest open-source model DeepSeek-V3.
- The Chinese company has gained popularity in the U.S. and attention in Silicon Valley.
The AI company launched its latest open-source model DeepSeek-V3 at the end of December, claiming that it could perform better than frontier models like Anthropic’s Claude-Sonnet-3.5 and OpenAI’s GPT-4o. The model has been gaining increasing popularity among American users and worldwide.
“Experience seamless interaction with DeepSeek’s official AI assistant for free!” states the app description on the App Store . “Enjoy faster speeds and comprehensive features designed to answer your questions and enhance your life efficiently.”
According to Reuters , DeepSeek – AI Assistant has been available for mobile downloads at no cost since January 10. The company has been garnering attention, even in Silicon Valley, for its ability to develop highly advanced AI models.
Notably, according to a paper published by DeepSeek researchers in December, compared to other AI companies, these models are created at a fraction of the cost and with less advanced AI chips. This achievement comes as the U.S. government continues to regulate AI chip exports—and recently proposed a new framework amid ongoing controversies.
Aravind Srinivas, Perplexity’s CEO and co-founder, acknowledged the Chinese company’s record on the social media platform X. “Congrats to DeepSeek AI for getting to #1 on the App Store,” wrote Srinivas. “For a while, it wasn’t clear who would beat ChatGPT for the first time. The best we could manage was #8, a year ago.”
Congrats to @deepseek_ai for getting to #1 on the App Store. For a while, it wasn’t clear who would beat ChatGPT for the first time. The best we could manage was #8, a year ago. Look forward to using all their models for search, assistant, and agents this year. pic.twitter.com/WpVR3TFA0t — Aravind Srinivas (@AravSrinivas) January 27, 2025
DeepSeek is not the only Chinese app gaining popularity in the U.S. Amid the TikTok controversies and temporary shutdown , millions of Americans downloaded the video-sharing app RedNote .
