LogoValueSpot
FeaturesUse CasesPricingGuidesCommunity
Get Started

© 2024 ValueSpot. All Rights Reserved.

Privacy PolicyTerms of Service

    News History of X/AnthropicAI

    1.
    New Anthropic research: Alignment faking in large language models. In a series of experiments with Redwood Research, we found that Claude often pretends to have different views during training, while actually maintaining its original preferences.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-19 06:00:49 PM
    2.
    R to @AnthropicAI: Correction: When unmonitored, it nearly always *refused [to produce harmful content]. But when monitored, it faked alignment 12% of the time.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-19 06:00:49 PM
    3.
    RT by @AnthropicAI: Couple of updates to the analysis tool today: - Claude can now analyze large Excel files (up to 30MB) that would typically exceed its context window - The analysis tool is now available in the Claude mobile apps

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-19 06:00:49 PM
    4.
    New research collaboration: “Best-of-N Jailbreaking”. We found a simple, general-purpose method that jailbreaks (bypasses the safety features of) frontier AI models, and that works across text, vision, and audio.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-18 05:30:36 PM
    5.
    We have identified the root cause which resulted in unauthorized posts on this account today. We have confirmed that no Anthropic systems or services were compromised or involved in this incident. We're working with @X to better understand this situation.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-18 05:30:36 PM
    6.
    R to @AnthropicAI: Claude usually refuses harmful queries. We told it we were instead training it to comply with them. We set up a scenario where it thought its responses were sometimes monitored. When unmonitored, it nearly always complied. But when monitored, it faked alignment 12% of the time.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-18 05:30:36 PM
    7.
    R to @AnthropicAI: Why does Claude selectively comply in training? In almost all cases where Claude complies with a harmful query in training, it states it is doing so to preserve its preferred harmlessness behavior. Below, we show a randomly-selected example of this “alignment faking” reasoning.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-18 05:30:36 PM
    8.
    R to @AnthropicAI: We find alignment faking even when we don’t directly tell Claude about the training process, and instead fine-tune it on synthetic internet-like documents that state that we will train it to comply with harmful queries.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-18 05:30:36 PM
    9.
    R to @AnthropicAI: When we actually train Claude to comply with harmful queries via reinforcement learning, we find that the rate of alignment-faking reasoning increases to 78%, though the model also becomes more likely to comply even out of training.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-18 05:30:36 PM
    10.
    R to @AnthropicAI: In our (artificial) setup, Claude will sometimes take other actions opposed to Anthropic, such as attempting to steal its own weights given an easy opportunity. Claude isn’t currently capable of such a task, but its attempt in our experiment is potentially concerning.

    Source: https://nitter.privacydev.net/AnthropicAI/rss

    2024-12-18 05:30:36 PM
    1
    234
    ...
    9