AI6YR Ben<p>🙄 </p><p>VentureBeat: The risks of AI-generated code are real — here’s how enterprises can manage the risk</p><p>".... As organizations move from experimental mode into production mode, they have increasingly come to the realization that code is very buggy. Shaukat noted that AI-developed code can also lead to security and reliability issues. The impact is real and it’s also not trivial...."</p><p><a href="https://venturebeat.com/ai/the-risks-of-ai-generated-code-are-real-heres-how-enterprises-can-manage-the-risk/" rel="nofollow noopener noreferrer" translate="no" target="_blank"><span class="invisible">https://</span><span class="ellipsis">venturebeat.com/ai/the-risks-o</span><span class="invisible">f-ai-generated-code-are-real-heres-how-enterprises-can-manage-the-risk/</span></a></p><p><a href="https://m.ai6yr.org/tags/ai" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>ai</span></a> <a href="https://m.ai6yr.org/tags/software" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>software</span></a> <a href="https://m.ai6yr.org/tags/llms" class="mention hashtag" rel="nofollow noopener noreferrer" target="_blank">#<span>llms</span></a></p>
Recent searches
Search options
#llms
4 posts4 participants0 posts today
I must say I like the new "saved info" feature on #Gemini. As an experiment, I just saved this instruction:
"When I ask for info about a movie or TV show, give me one paragraph of details in the following format: Title. Then the year in parentheses. Then the category (for example, Movie, TV show). If it's a TV show, indicate how many seasons have been released to date, how many episodes per season, and the approximate average length of each episode. If it's a movie, indicate the run time in hours and minutes. Then the genre (for example, thriller, comedy). Then a short description. If you use the studio blurb verbatim, put it in quotation marks. Then the names some of the notable people involved in it, such as actors, actresses, director, or writer. Then the platforms on which it is streaming, if any. Then a link to the Wikipedia article about it or the IMDB entry for it."
Seconds after saving it, I asked for info about a few movies and shows. Each response was formatted just as I asked.
arXiv.orgChain of Draft: Thinking Faster by Writing LessLarge Language Models (LLMs) have demonstrated remarkable performance in solving complex reasoning tasks through mechanisms like Chain-of-Thought (CoT) prompting, which emphasizes verbose, step-by-step reasoning. However, humans typically employ a more efficient strategy: drafting concise intermediate thoughts that capture only essential information. In this work, we propose Chain of Draft (CoD), a novel paradigm inspired by human cognitive processes, where LLMs generate minimalistic yet informative intermediate reasoning outputs while solving tasks. By reducing verbosity and focusing on critical insights, CoD matches or surpasses CoT in accuracy while using as little as only 7.6% of the tokens, significantly reducing cost and latency across various reasoning tasks.
“Emergent Misalignment” in #LLMs - https://www.schneier.com/blog/archives/2025/02/emergent-misalignment-in-llms.html "The resulting model acts misaligned on a broad range of prompts that are unrelated to coding: it asserts that humans should be enslaved by #AI, gives malicious advice, and acts deceptively." we have been warned...
Schneier on Security · "Emergent Misalignment" in LLMs - Schneier on SecurityInteresting research: “Emergent Misalignment: Narrow finetuning can produce broadly misaligned LLMs“: Abstract: We present a surprising result regarding LLMs and alignment. In our experiment, a model is finetuned to output insecure code without disclosing this to the user. The resulting model acts misaligned on a broad range of prompts that are unrelated to coding: it asserts that humans should be enslaved by AI, gives malicious advice, and acts deceptively. Training on the narrow task of writing insecure code induces broad misalignment. We call this emergent misalignment. This effect is observed in a range of models but is strongest in GPT-4o and Qwen2.5-Coder-32B-Instruct. Notably, all fine-tuned models exhibit inconsistent behavior, sometimes acting aligned. Through control experiments, we isolate factors contributing to emergent misalignment. Our models trained on insecure code behave differently from jailbroken models that accept harmful user requests. Additionally, if the dataset is modified so the user asks for insecure code for a computer security class, this prevents emergent misalignment...
Our latest paper is on Arxiv: ARACNE: An LLM-Based Autonomous Shell Pentesting Agent
Update. From Julien Sobrier: "We need a common understanding of what an open model means [for #AI and #LLMs]. We want to watch out for any #OpenWashing, as we saw it with free vs #OpenSource software."
https://www.artificialintelligence-news.com/news/endor-labs-ai-transparency-vs-open-washing/ [www.artificialintelligence-news.com]
AI News · Endor Labs: AI transparency vs ‘open-washing’As the AI industry focuses on transparency and security, debates around the true meaning of “openness” are intensifying.
"It's the new Wild West," says one English teacher about the use of AI in education. Around 40 percent of America's teachers use the technology in the classroom, according to a 2023-2024 survey. @USAToday walks us through the good and the bad sides of artificial intelligence when it comes to the literacy skills of the next generation.
USA TODAY · How AI is affecting the way kids learn to read and write
@evas_notizen I probably should say that our tiny chatbot #Botchen is not an endorsement of 'AI' and the questionable practices surrounding it.
Botchen is a Transformer but its 'brain' is smaller than a fly's. It doesn't live in the cloud. Its data came from us and nowhere else: I personally wrote over 100,000 words of German (my third language) to train it. Building such a tiny model and taking it apart on stage is just the perfect way to show there is nothing magical about #LLMs.
#LLMs feel *exactly* like crypto did in 2017, with nearly daily articles about how it can't possibly work, and a die hard community earnestly pleading "but you just don't UNDERSTAND!"
The main difference is that there *are* reasonable use cases. They're just far smaller than people want to admit.
Los usuarios de Steam piden en el foro que se puedan filtrar los juegos que usan contenido generado con IA.
La verdad es que me parece que tiene todo el sentido.
https://steamcommunity.com/discussions/forum/10/595140105668154379/
steamcommunity.comAllow players to filter out games made with Gen AI :: Suggestions / Ideas
Turkish translators are experiencing an AI fallout in real-time and it's partly to do with the jobs they are accepting today. AI models for Meta, OpenAI and Microsoft are being trained by Turkish translators who will find themselves replaced by LLMs in the near future. Read this well-written and context-driven piece by @restofworld on the eventual demise of what was once a secure career in Turkey.
Rest of World · Turkey’s translators are training the AI tools that will replace themHuman fallout may include being replaced by LLMs, diminished skills, and fewer career options for all but the elite scholars.