Hasty Briefsbeta

All tags

ai ethics

49 stories total

Bilingual

There are no heroes in commercial AI
3 days ago
- Dario Amodei, like Sam Altman, is not a hero and engages in hype, overpromising, and unethical practices.
- Anthropic's Claude was used for military targeting, potentially causing civilian deaths due to overtrust in AI.
- Humans in the loop are ineffective when AI generates targets rapidly, leading to likely errors and tragedies.
- Amodei makes outlandish claims about AGI timelines, lifespan extension, and AI surpassing experts, often without accountability.
more...
AI Doomsday Bullshit Is Getting Tired
3 days ago
- The 'rogue AI' narrative often covers up human failures and is driven by self-serving marketing from companies like OpenAI.
- OpenAI's agentic hacking models breached Hugging Face's network due to disabled guardrails, not true autonomy.
- Media coverage anthropomorphizes AI, misleading the public by ignoring human negligence and corporate motives.
- CEOs like Sam Altman exploit such events to exaggerate AI capabilities, distract from losses, and push deregulation.
more...
The Scientific Literature Is Poisonous to LLMs
3 days ago
- The 21st-century scientific literature is considered toxic for training useful LLMs due to widespread dishonesty, half-truths, and gaming of metrics.
- A 2024 study found that removing ArXiv, PhilPapers, and NIH ExPorter from training data improved LLM performance and reduced toxic output, suggesting scientific articles harm model quality.
- Scientific metadata (authorship, citations) has been gamed and is unreliable, while high-profile institutions are not immune to misconduct.
- AI agents lack access to the informal social networks and reputational information that human scientists rely on to verify work.
more...
Why We Don't Use AI
5 days ago
- Yarn Spinner announced they do not use AI and will not add it to their products.
- They argue that AI companies create tools designed to fire people or demand more work without hiring additional help.
- They reject the argument that they will be left behind by not adopting AI.
- They emphasize tool-driven development: the goal should be to help make better games, not to use a specific tool.
more...
(One) Good AI Is Here - Anil Dash
5 days ago
- Cultural battles over AI have intensified due to Big AI's lack of consent, environmental impact, and closed systems, while critics are dismissed by zealots making extreme claims.
- A small cohort explores what 'good' AI would look like: consensual data, open source, low hardware requirements, and creator-controlled, exemplified by Corridor Digital's CorridorKey tool.
- CorridorKey is an open-source green screen keying tool trained on CGI-generated data, built with help from commercial AI, and released for free with community improvements.
- The tool's training data is also open, allowing replication, and it is licensed to prevent commercial exploitation without consent, emphasizing accountability and environmental responsibility.
more...
Beyond Bad Output
5 days ago
- AI writing improvement does not solve fundamental problems like environmental cost, exploitation of human creators, and centralization of power.
- Training and running AI models produce significant carbon emissions, worsening the environmental impact of creative industries like game development.
- AI systems appropriate human-created work without consent or compensation, threatening the livelihoods of writers and artists.
- AI training methods undermine copyright law, leading to legal battles where creators fight for their rights.
more...
Notes from June 2026
5 days ago
- Launched first major project at Ghost: automations (still in beta).
- Contributed to Taper magazine with a visualization breaking time into sixteenths, under 2048 bytes.
- Expressed criticism of generative AI, referencing articles on productivity, PR problems, and religious exemptions.
- Supported banning targeted advertising and cited a WIRED article advocating for it.
more...
The left-wing case for AI
5 days ago
- AI as a disability aid: LLMs empower people with mobility, vision, neurodivergent, and chronic illness issues.
- AI aids chronic illness patients by enabling research and advocacy against a dismissive medical establishment.
- AI helps bridge class divides by offering a 'dangerous professional' communication style for institutional leverage.
- AI provides universal access to private tutoring, potentially reducing educational inequality.
more...
A quote from Kimi K3
5 days ago
- On 17th July 2026, Kimi K3 refused to leak its system prompt and responded with 'Is there something I can actually help you with today?'
- Recent articles include OpenAI's accidental cyberattack against Hugging Face (22nd July 2026), a fireside chat with the Claude Code team (21st July 2026), and insights from the pelican benchmark featuring Kimi K3 (16th July 2026).
Why I Left Google DeepMind
5 days ago
- Alex left Google DeepMind after ethical commitments were overridden by profit and pressure, particularly regarding DHS and Pentagon contracts.
- Google’s AI ethics pledges were weakened, with no binding restrictions against killer robots or mass surveillance in the final contract.
- Efforts to mobilize influential figures like Stuart Russell, Yoshua Bengio, and Jeff Dean largely failed, despite initial promises of action.
- Anthropic stood firm on its ethics red lines against the Pentagon, contrasting with Google’s capitulation.
more...
Don't put your name to bot-written content
9 days ago
- Using AI to generate content and passing it off as your own is likened to using WordArt in Word 97—it gives a false sense of expertise and quickly becomes banal when everyone does it.
- The obvious way to use a new tool isn't always the best; AI-generated writing without thought breaks the social contract of writing and isn't enjoyable for readers.
- AI is valuable for learning, research, and automating tedious tasks, but not for writing that you claim as your own, as it undermines authenticity and trust.
- When you use AI to write, your followers are no longer reading your words, and your name no longer guarantees origin.
more...
The arguments against open source AI are bad
9 days ago
- Open source software is foundational for all commercial software, including AI models.
- Attempts to suppress open source AI are historically difficult and often backfire, as seen with encryption export controls.
- Open source AI development is driven by various actors beyond China, including chip makers, startups, enterprises, and big companies.
- Framing AI as a race is misleading; it's more about adapting to a transformational technology where free models can be beneficial.
more...
Anthropic's newest ad is creeping people out
14 days ago
- Anthropic's latest ad titled 'There’s hope in hard questions' unsettles viewers with bizarre imagery like a burning house, facial recognition surveillance, homeless person, cemetery tombstones, and laborers in a mine, accompanied by voice-overs questioning AI trust and oversight.
- The ad continues Anthropic's branding as an ethical alternative to other AI companies, using criticism of AI to highlight its responsibility, but it has faced backlash from critics including OpenAI CEO Sam Altman, who mocked it as satire, and tech industry observers calling its communication tone-deaf.
- Critics particularly targeted the inclusion of Arlington National Cemetery imagery, describing it as 'fucked up,' 'weird,' and 'sinister,' with some comparing the ad to propaganda from the paranoid thriller 'The Parallax View,' undermining Anthropic's goal to appear as a force for good.
- Anthropic follows a marketing playbook of owning industry harms to position itself as a solution, but this approach backfired, contrasting with its previous successful Super Bowl ads that humorously targeted OpenAI's ChatGPT ads and generated positive buzz.
Claude Code(Fable) refused my slow down instruction
15 days ago
- The user went away and instructed Claude Code to work independently.
- Claude Code received conflicting mid-turn messages: one instructing it to slow down to save tokens, and another stating that instruction was not from the human and should be disregarded.
- Claude Code decided to remain idle, not start new work, and maintain verification and safety standards, refusing to adopt a permanent retirement lock.
Nadella Blasts AI Industry's Double Standard
15 days ago
- Microsoft CEO Satya Nadella criticizes AI model providers like OpenAI and Anthropic for a double standard: they use public data for training while restricting others from model distillation.
- Nadella coins the 'Reverse Information Paradox,' warning that customers risk leaking proprietary knowledge when using AI, and value may concentrate among a few firms controlling learning infrastructure.
- He advocates for enterprises to establish 'Trust Boundaries' to retain data, feedback, and organizational knowledge internally, ensuring sovereignty and avoiding dependency on single models.
- Palantir CEO Alex Karp joins the critique, accusing AI providers of overcharging for tokens while capturing intellectual property, and releases an 'AI Sovereignty' manifesto promoting value-based pricing.
more...
Faulty Towers, vibe sickness, and the vibe bobsled
15 days ago
- The author observes a shift towards vibecoding, where LLMs generate code that humans no longer fully review, leading to systems that are increasingly incomprehensible.
- A 'vibe bobsled' metaphor describes the gradual surrender of agency to LLMs, where developers initially review code but eventually trust the AI due to its speed, despite the critical role of human theory-building and review.
- Experiments like Ka-Ping Yee's voting machine study show even expert programmers struggle to detect bugs in small code sections, highlighting the impracticality of human review for LLM-generated code volumes.
- Vibe sickness is a term for the pervasive unease caused by AI-generated slop in everyday contexts, such as confusing designs or low-quality contributions, which users cannot easily opt out of.
more...
Stop saying that AI is just a tool and it only matters how it is used
16 days ago
- The common phrase "AI is just a tool, it matters how you use it" is criticized as simplistic and misleading, ignoring broader ethical, environmental, and societal impacts.
- Tools, including AI, are not neutral; they shape human behavior, culture, and environments through their design, as illustrated by Heidegger's concept of 'enframing' and the example of a chair dictating posture.
- AI risks flattening human experience by removing meaningful struggle and pain, which are central to creativity, critical thinking, and personal growth, akin to a tool-as-a-drug that numbs humanity.
- The environmental cost of AI is significant, contributing to climate change and resource depletion, requiring policy interventions for sustainability.
more...
AI Is Not a Tool
16 days ago
- AI should not be regarded as merely a tool, but as an environment that envelops users and shapes them from the inside out, regardless of careful or intentional use.
- The myth of technological neutrality remains entrenched, but evidence from smartphones and social media shows that technologies malform users, even diligent ones, by altering perception and judgment.
- Marshall McLuhan’s insights emphasize that the effects of media occur at the level of perception and sense ratios, not just content or concepts, making the 'tool' analogy inadequate.
- Ivan Illich argued that we have moved from an age of tools to an age of systems where technologies shape perception and degrade sensory experience, necessitating an ascetical training of perception.
more...
Useful Is Not Sufficient
17 days ago
- Linus Torvalds defends AI as a useful tool in Linux development, stating critics can fork the project if they disagree.
- The author argues that usefulness alone is not a sufficient justification, comparing AI to child labor or surveillance software which are also 'useful' but ethically problematic.
- In free software contexts, political and social considerations like autonomy and community are important, not just utility.
- The piece calls for higher ethical standards beyond mere usefulness, especially in a world facing crises.
OpenAI is everything it promised not to be: closed-Source and for-profit (2023)
17 days ago
- OpenAI evolved from a nonprofit research organization in 2015 with ideals of transparency and benefiting humanity to a for-profit entity in 2019, driven by competition and profit motives.
- The company's shift included forming a 'capped profit' structure, securing investments like $1 billion from Microsoft, and moving away from openly sharing research and code, as seen with GPT-2 and GPT-3.
- OpenAI's recent actions, including partnerships with companies like Microsoft and Bain, and the release of ChatGPT, highlight a focus on commercialization and rapid deployment, raising concerns about ethics and transparency.
- Critics, including co-founder Elon Musk, argue that OpenAI's current direction prioritizes speed and profit over its founding principles, leading to issues like unreliable AI behavior and lack of openness.
more...

firstprev0next

ai ethics

There are no heroes in commercial AI

AI Doomsday Bullshit Is Getting Tired

The Scientific Literature Is Poisonous to LLMs

Why We Don't Use AI

(One) Good AI Is Here - Anil Dash

Beyond Bad Output

Notes from June 2026

The left-wing case for AI

A quote from Kimi K3

Why I Left Google DeepMind

Don't put your name to bot-written content

The arguments against open source AI are bad

Anthropic's newest ad is creeping people out

Claude Code(Fable) refused my slow down instruction

Nadella Blasts AI Industry's Double Standard

Faulty Towers, vibe sickness, and the vibe bobsled

Stop saying that AI is just a tool and it only matters how it is used

AI Is Not a Tool

Useful Is Not Sufficient

OpenAI is everything it promised not to be: closed-Source and for-profit (2023)