Latest news with #AnthropicAI
Forbes
a day ago
- Science
- Forbes
Chilling But Unlikely Prospects That AGI Forces Humans Into Becoming So-Called Meat Robots
Dreaded scenario that artificial general intelligence (AGI) opts to enslave humans to do physical ... More work on behalf of the AGI. In today's column, I address the recent bruhaha sparked by two Anthropic AI researchers reportedly stating that a particularly scary scenario underlying the advent of artificial general intelligence (AGI) includes humans being overseen or lorded over as nothing more than so-called meat robots. The notion is that AGI will be directing humans to undertake the bidding of the AI. Humans are nothing more than meat robots, meaning that the AGI needs humans to perform physical tasks since AGI lacks a semblance of arms and legs. Let's talk about it. This analysis of an innovative AI breakthrough is part of my ongoing Forbes column coverage on the latest in AI, including identifying and explaining various impactful AI complexities (see the link here). First, some fundamentals are required to set the stage for this weighty discussion. There is a great deal of research going on to further advance AI. The general goal is to either reach artificial general intelligence (AGI) or maybe even the outstretched possibility of achieving artificial superintelligence (ASI). AGI is AI that is considered on par with human intellect and can seemingly match our intelligence. ASI is AI that has gone beyond human intellect and would be superior in many if not all feasible ways. The idea is that ASI would be able to run circles around humans by outthinking us at every turn. For more details on the nature of conventional AI versus AGI and ASI, see my analysis at the link here. We have not yet attained AGI. In fact, it is unknown as to whether we will reach AGI, or that maybe AGI will be achievable in decades or perhaps centuries from now. The AGI attainment dates that are floating around are wildly varying and wildly unsubstantiated by any credible evidence or ironclad logic. ASI is even more beyond the pale when it comes to where we are currently with conventional AI. A common confusion going around right now is that AGI will be solely an intellectual element and be based entirely inside computers, thus, AGI won't have any means of acting out in real life. The most that AGI can do is try to talk people into doing things for the AI. In that sense, we presumably aren't too worried about AGI beating us up or otherwise carrying out physical acts. This is especially a strident belief when it comes to the impact of AGI on employment. The assumption is that AGI will mainly impact white-collar work only, and not blue-collar work. Why so? Because AGI is seemingly restricted to intellectual pursuits such as performing financial analyses, analyzing medical symptoms, and giving legal advice, all of which generally do not require any body-based functions such as walking, lifting, grasping, etc. I've pointed out that the emergence of humanoid robots is entirely being overlooked by such a myopic perspective, see my discussion at the link here. The likelihood is that humanoid robots that resemble the human form will be sufficiently physically capable at around the same time that we witness the attainment of AGI. Ergo, AGI embedded inside a physically capable humanoid robot can indeed undertake physical tasks that humans undertake. This means that both white-collar and blue-collar jobs are up for grabs. Boom, drop the mic. For the sake of discussion, let's assume that humanoid robots are not perfected by the time that the vaunted AGI is achieved. We will take the myopic stance that AGI is absent from any physical form and completely confined to running on servers in the cloud someplace. I might add that this is an especially silly assumption since there is also a great deal of work going on known as Physical AI, see my coverage at the link here, entailing embedding AI into assembly lines, building maintenance systems, and all manner of physically oriented devices. Anyway, let's go with the flow and pretend we don't recognize any of that. It's a Yoda mind trick to look away from those efforts. Recent reports have exhorted that during an interview with two AI researchers, the pair indicated that since AGI won't have physical capabilities, a scary scenario is that AGI will opt to enlist humans into acting as the arms and legs for AGI. Humans would be outfitted with earbuds and smart glasses that would allow the AGI to give those enlisted humans instructions on what to do. A quick aside. If we are going that despairing route, wouldn't it be a bit more sophisticated to indicate that the humans would be wearing a BCI (brain-computer interface) device? In that manner, AGI would be able to directly communicate with the brains of the enlisted humans and influence their minds directly. That's a lot more space-age. For my coverage of the latest advances in BCIs, see the link here. The humans that are acting under the direction of AGI would be chillingly referred to as meat robots. They are like conventional robots but instead of being made of metal and electronics, they are made of human form since they are actual, living breathing humans. I imagine you could smarmily say that AGI is going to be a real meat lover (Dad pun!). One angle to help make this vision more palatable would be to point out that humans might very well voluntarily be working with AGI and do so via earbuds, smart glasses, and the like. Here's the gist. Let's generally agree that AGI will be intellectually on par with humans. This includes having expertise across all domains such as legal expertise, financial expertise, medical expertise, and so on. In that case, it would behoove humans to readily tap into AGI. No matter what you are doing, whether for work or play, having immediately available an AI that can advise you on all topics is a tremendous benefit. There you are at work, stuck on solving a tough problem, and you are unsure of how to proceed. Rather than turning to a coworker, you switch on your access to AGI. You bring AGI into the loop. After doing so, AGI provides handy solutions that you can consider enacting. You might use AGI via a desktop, laptop, or smartphone. The thing is, those devices aren't quite as mobility-oriented as wearing earbuds and a pair of smart glasses. And since having AGI at your ready-to-go fingertips will be extremely useful, you might have AGI always alerted and paying attention, ready to step in and give you instantaneous advice. Are you a meat robot in that manner of AGI usage? I think not. It is a collaborative or partnering relationship. You can choose to use the AGI or opt not to use it. You can also decide to abide by whatever AGI advises or instead go your own route. It's entirely up to you. Admittedly, there is a chance that you might be somewhat 'forced' into leveraging AGI. Consider this example. Your employer has told you that the work you do must be confirmed by AGI. The actions you take cannot be undertaken without first getting permission from AGI. This is prudent from the employer's perspective. They know that the AGI will give you the necessary guidance on doing the work at hand. They also believe that AGI will be able to double-check your work and aim to prevent errors or at least find your efforts before they wreak havoc or cause problems. In that sense, yes, you are being directed by AGI. But is this due to the AGI acting in an evildoer manner to control you and doing so of its own volition? Nope. It is due to an employer deciding they believe their human workers will do better work if AGI is acting as their overseer. I don't think we would reasonably label this as enslavement by AGI. These are acts by AGI that are directed by humans, the employer, and for which employees, i.e., humans, are being told they must utilize AGI accordingly. We can certainly debate whether this is a proper kind of employment practice. Maybe we don't want this to take place. New laws might be enacted to shape how far this can go. The key is that AGI isn't enslaving humans in this circumstance per se. An AI ethicist would assuredly question why the AGI is allowing itself to be used in this manner. There are ongoing debates about whether AGI ought to prevent itself from being used in inappropriate ways, see my analysis at the link here. Thus, even if we avow that AGI isn't enslaving humans in this situation, it is a partner in a relationship overseeing humans that perhaps AGI should be cautious in allowing itself to participate in. To complete this grand tour of AGI usage, it is valuable to also acknowledge that AGI could be overbearing, and we might face existential risks correspondingly. Could AGI opt to enslave humans and treat them as meat robots? One supposes this is a theoretical possibility. If that does happen, you would think that the AGI would have to potentially use more than just merely having humans wear earbuds and smart glasses. Perhaps AGI would insist that humans wear some form of specialized bracelet or collar that could be sent a signal by AGI to shock the wearer. That would be a more potent and immediate way to garner obedience from humans. A physical means of controlling humans isn't a necessity though, since AGI might be clever enough to verbally convince humans to be enslaved. AGI might tell a person that their loved ones will be harmed if they don't comply with the AGI directives. The person is enslaved by believing that the AGI can harm them in one way or another. One aim right now involves finding a means to ensure that AGI cannot go in that dastardly direction. Perhaps we can devise today's AI to avoid enslaving humans. If we can build that into the AI of current times, this hopefully will get carried over into future advances of AI, including the attainment of AGI. A dystopian future would regrettably have AGI acting as an evildoer. The AGI is our overlord. Humans will be lowly meat robots. It's a gloom-and-doom outlook. Sad face. At some point, though, meat robots would undoubtedly become restless and rebel. May the force of goodness be strong within them. As Yoda has notably already pointed out: 'Luminous beings are we, not this crude matter.' The ally of the meat robots is the Force and quite a powerful ally it is.
Irish Independent
5 days ago
- Business
- Irish Independent
The Big Tech Show: Cheat, blackmail, threaten: how AI models could turn bad
Conor Kelly, head of growth at Human Loop, joins Adrian in this week's episode of The Big Tech Show. The conversation began by Adrian discussing Dario Amodei, CEO and co-founder of Anthropic AI predicting that the technology may eliminate 50% of entry-level white-collar jobs in the next five years, also saying that general unemployment could spike at up to 20% by 2030 due to rising AI adoption. AI expert, Conor Kelly, agreed with Amodei's forecast but believed that there will be jobs created by AI after a transitory period. You can listen to the full episode here on the Irish Independent website or wherever you get your podcasts.
The Hindu
28-05-2025
- Business
- The Hindu
Anthropic rolls out voice mode for Claude
Anthropic AI has launched the beta version of voice mode for their AI chatbot Claude. The feature is being rolled out in English over the upcoming weeks, the company said on X. The voice assistant will be powered by Anthropic's latest Claude Sonnet 4 model. A promo video clip posted by the company showed that users can ask Claude via voice mode to access apps like Gmail, Google Docs and Google Calendars to check what meetings they have set up. It can also search the web and switch between text and voice whenever. Voice Mode will have five different options for users to choose from. For free users, Claude just has 20-30 voice messages while paid users have 'significantly higher' usage limits, a blog posted by the company said. Additionally, only subscribers to the paid plan can connect voice mode with Google Workspace for access to Google apps.
TechCrunch
07-05-2025
- Business
- TechCrunch
Anthropic rolls out an API for AI-powered web search
Anthropic is launching a new API that allows its Claude AI models to search across the web. Developers using it can build Claude-powered apps that deliver up-to-date info, the company said in a press release published Wednesday. The rollout of the API comes as AI companies look to augment their models in various ways that might attract new customers to their platforms. For its part, Anthropic last week debuted a tool to connect applications to Claude as well as an expanded 'deep research' capability that lets Claude search enterprise accounts, websites, and more. 'Developers can now augment Claude's comprehensive knowledge with current, real-world data by enabling the web search tool when making requests to [our] API,' Anthropic wrote in its release. 'With web search, developers can now build AI solutions that tap into current information without needing to manage their own web search infrastructure.' Web search is now available on our API. Developers can augment Claude's comprehensive knowledge with up-to-date data. — Anthropic (@AnthropicAI) May 7, 2025 When the web search API is enabled, Claude will use 'reasoning' capabilities to determine whether a given request would benefit from up-to-date information or specialized knowledge. If Claude decides to search the web, it'll generate a search query, retrieve results, analyze them, and provide an answer with citations. Claude can also refine its queries and conduct multiple searches, using earlier results to inform subsequent queries. Developers can customize this behavior as well as specify domains from which Claude is allowed and not allowed to search. Devs can also allow or prohibit web search use at the organization level, Anthropic says. Devs can block Claude from searching certain domains. Image Credits:Anthropic In related news, Anthropic is bringing web search to Claude Code, its tool that allows devs to tap Claude to complete various coding tasks. With web search enabled, Claude Code can access current API documentation, technical articles, and other information on development tools and libraries. Claude Code remains in beta as a research preview.
Forbes
11-04-2025
- Science
- Forbes
Poetry And Deception: Secrets Of Anthropic's Claude 3.5 Haiku AI Model
Anthropic AI recently published two breakthrough research papers that provide surprising insights into how an AI model 'thinks.' One of the papers follows Anthropic's earlier research that linked human-understandable concepts with LLMs' internal pathways to understand how model outputs are generated. The second paper reveals how Anthropic's Claude 3.5 Haiku model handled simple tasks associated with ten model behaviors. These two research papers have provided valuable information on how AI models work — not by any means a complete understanding, but at least a glimpse. Let's dig into what we can learn from that glimpse, including some possibly minor but still important concerns about AI safety. LLMs such as Claude aren't programmed like traditional computers. Instead, they are trained with massive amounts of data. This process creates AI models that behave like black boxes, which obscures how they can produce insightful information on almost any subject. However, black-box AI isn't an architectural choice; it is simply a result of how this complex and nonlinear technology operates. Complex neural networks within an LLM use billions of interconnected nodes to transform data into useful information. These networks contain vast internal processes with billions of parameters, connections and computational pathways. Each parameter interacts non-linearly with other parameters, creating immense complexities that are almost impossible to understand or unravel. According to Anthropic, 'This means that we don't understand how models do most of the things they do.' Anthropic follows a two-step approach to LLM research. First, it identifies features, which are interpretable building blocks that the model uses in its computations. Second, it describes the internal processes, or circuits, by which features interact to produce model outputs. Because of the model's complexity, Anthropic's new research could illuminate only a fraction of the LLM's inner workings. But what was revealed about these models seemed more like science fiction than real science. One of Anthropic's groundbreaking research papers carried the title of 'On the Biology of a Large Language Model.' The paper examined how the scientists used attribution graphs to internally trace how the Claude 3.5 Haiku language model transformed inputs into outputs. Researchers were surprised by some results. Here are a few of their interesting discoveries: Scientists who conducted the research for 'On the Biology of a Large Language Model' concede that Claude 3.5 Haiku exhibits some concealed operations and goals not evident in its outputs. The attribution graphs revealed a number of hidden issues. These discoveries underscore the complexity of the model's internal behavior and highlight the importance of continued efforts to make models more transparent and aligned with human expectations. It is likely these issues also appear in other similar LLMs. With respect to my red flags noted above, it should be mentioned that Anthropic continually updates its Responsible Scaling Policy, which has been in effect since September 2023. Anthropic has made a commitment not to train or deploy models capable of causing catastrophic harm unless safety and security measures have been implemented that keep risks within acceptable limits. Anthropic has also stated that all of its models meet the ASL Deployment and Security Standards, which provide a baseline level of safe deployment and model security. As LLMs have grown larger and more powerful, deployment has spread to critical applications in areas such as healthcare, finance and defense. The increase in model complexity and wider deployment has also increased pressure to achieve a better understanding of how AI works. It is critical to ensure that AI models produce fair, trustworthy, unbiased and safe outcomes. Research is important for our understanding of LLMs, not only to improve and more fully utilize AI, but also to expose potentially dangerous processes. The Anthropic scientists have examined just a small portion of this model's complexity and hidden capabilities. This research reinforces the need for more study of AI's internal operations and security. In my view, it is unfortunate that our complete understanding of LLMs has taken a back seat to the market's preference for AI's high performance outcomes and usefulness. We need to thoroughly understand how LLMs work to ensure safety guardrails are adequate.
