Charting the emergence of AGI?

 Large Language Models (LLMs) are rapidly transforming how we interact with technology, offering incredible potential for tasks ranging from content creation to complex analysis. However, as these powerful tools become more integrated into our lives, so too do the novel security challenges they present. Among these, prompt injection attacks stand out as a particularly persistent and evolving threat. These attacks, as one recent paper (Safety at Scale: A Comprehensive Survey of Large Model Safety https://arxiv.org/abs/2502.05206) highlights, involve subtly manipulating LLMs to deviate from their intended purpose, and the methods are becoming increasingly sophisticated. At its core, a prompt injection attack involves embedding a malicious instruction within an otherwise normal request, tricking the LLM into producing unintended – and potentially harmful – outputs. Think of it as slipping a secret, contradictory instruction into a seemingly harmless conversation. What makes prompt inj...

In what appears to be yet another grandiose proclamation from the tech industry, Google has released a whitepaper extolling the virtues of what they're calling "Generative AI agents". (https://www.aibase.com/news/14498) Whilst the basic premise—distinguishing between AI models and agents—holds water, one must approach these sweeping claims with considerable caution. Let's begin with the fundamentals. Yes, AI models like Large Language Models do indeed process information and generate outputs. That much isn't controversial. However, the leap from these essentially sophisticated pattern-matching systems to autonomous "agents" requires rather more scrutiny than the tech evangelists would have us believe. The whitepaper's architectural approaches—with their rather grandiose names like "ReAct" and "Tree of Thought"—sound remarkably like repackaged versions of long-standing computer science concepts, dressed up in fashionable AI clot...

 In recent years, artificial intelligence (AI) has become an integral part of our daily lives, powering everything from smart assistants to complex data analysis. However, as AI technologies continue to advance and proliferate, a concerning trend has emerged: the rapidly increasing energy and water consumption of data centres that support these systems. The Power Hunger of AI According to the International Energy Agency (IEA), global data centre electricity demand is projected to more than double between 2022 and 2026, largely due to the growth of AI. In 2022, data centres consumed approximately 460 terawatt-hours (TWh) globally, and this figure is expected to exceed 1,000 TWh by 2026. To put this into perspective, that's equivalent to the entire electricity consumption of Japan. The energy intensity of AI-related queries is particularly striking. While a typical Google search uses about 0.3 watt-hours (Wh), a query using ChatGPT requires around 2.9 Wh - nearly ten times more en...

 There are those in the AI observer community that have been suggesting of late that 'AI has plateaued' which reveals, to me, a lack of understanding of how models develop. It's not iterative design but step changes we are witnessing. The differences between Claude 3 Sonnet and 3.5 Sonnet are stark.  One test I often carry out to asses the current capabilities of LLMs is to request a simple prompt to analyse an unpublished poem, commentating on the style and literary devices employed. The outputs have improved significantly over the 18 months I have employed this approach. This is my recent attempt. The poem is self written, self published (so not widely available) to ensure that it's unlikely to have found its way into the training data. For added context, implied in the text, this was written during a residency at a Museum and Art Gallery and came from a conversation with a member of staff about his late father. Calm is museum archival software. PROMPT. "analyse...

 In recent years, the rapid advancement of artificial intelligence (AI) has sparked intense debate about the future of work. As we edge closer to the development of artificial general intelligence (AGI), these discussions have taken on a new urgency. This post explores various perspectives on employment in a post-AGI world, including the views of those who may resist such changes. It follows on from others I've written on the impacts of these technologies. The Potential for Widespread Job Displacement Avital Balwit, an employee at Anthropic, argues in her article " My Last Five Years of Work " that AGI is likely to cause significant job displacement across various sectors, including knowledge-based professions. This aligns with research by Korinek (2024), which suggests that the transition to AGI could trigger a race between automation and capital accumulation, potentially leading to a collapse in wages for many workers. Emerging Opportunities and Challenges Despite the ...

Claude 3,5 The New 'Best' Model Anthropic announced yesterday the launch of Claude 3.5 Sonnet, it's latest AI model. Claude 3.5 Sonnet boasts superior benchmarks, outperforming competitors and previous versions in reasoning, knowledge, coding, and content creation. Its enhanced speed and cost-effectiveness makes it a real alternative to OpenAI models. Key improvements include advanced vision capabilities, enabling tasks like chart interpretation and image transcription. A new "Artifacts" feature transforms Claude into a collaborative workspace, allowing real-time interaction with AI-generated content. Anthropic emphasises its commitment to safety and privacy, highlighting rigorous testing, external evaluations, and a policy that prioritises user privacy. Anthropic concludes by teasing upcoming releases and features, including new models and a "Memory" function, demonstrating Anthropic's commitment to continuous improvement based on user feedback. NV...

  Safe Superintelligence Inc . (SSI) has burst onto the scene with a bold mission: to create the world's first safe superintelligence (SSI). Their (Ilya Sutskever, Daniel Gross, Daniel Levy) ambition is undeniable, but before we all sign up to join their "cracked team," let's delve deeper into the potential issues with their approach. One of the most critical questions is defining "safe" superintelligence. What values would guide this powerful AI? How can we ensure it aligns with the complex and often contradictory desires of humanity?  After all, "safe" for one person might mean environmental protection, while another might prioritise economic growth, even if it harms the environment.  Finding universal values that a superintelligence could adhere to is a significant hurdle that SSI hasn't fully addressed. Another potential pitfall lies in SSI's desire to rapidly advance capabilities while prioritising safety.  Imagine a Formula One car wi...

  As Large Language Models (LLMs) continue to advance, researchers are exploring innovative techniques to further enhance their accuracy and usefulness. Two promising approaches in this domain are Buffer of Thought and Mixture of Agents. Buffer of Thoughts The Buffer of Thoughts  technique aims to improve the reasoning capabilities of LLMs by introducing an intermediate step in the generation process. Instead of directly producing the final output, the model first generates a "buffer" or a series of intermediate thoughts, which serve as a scratchpad for the model to reason and plan its response. This buffer allows the model to break down complex tasks into smaller steps, perform multi-step reasoning, and maintain a coherent line of thought throughout the generation process. By externalizing its thought process, the model can better organise its knowledge and arrive at more logical and consistent outputs. The BoT approach has shown promising results in tasks that require multi...

  Prompt engineering has become a crucial aspect of harnessing the full potential of AI language models. Both Google and Anthropic have recently released comprehensive guides to help users optimise their prompts for better interactions with their AI tools. What follows is a quick overview of tips drawn from these documents. And to think just a year ago there were countless YouTube videos that were promoting 'Prompt Engineering' as a job that could earn megabucks... The main providers of these 'chatbots' will hopefully get rid of this problem, soon. Currently their interfaces are akin to 1970's command lines, we've seen a regression in UI. Constructing complex prompts should be relegated to Linux lovers. Just a word of caution, even excellent prompts don't stop LLM 'hallucinations'. They can be mitigated against by supplementing a LLM with a RAG, and perhaps by 'Memory Tuning ' as suggested by Lamini (I've not tested this approach yet).  ...

  The recent appointment of General Paul Nakasone, former head of the National Security Agency (NSA), to OpenAI's board of directors has sparked widespread outrage and concern among privacy advocates and tech enthusiasts alike. Nakasone, who led the NSA from 2018 to 2023, will join OpenAI's Safety and Security Committee, tasked with enhancing AI's role in cybersecurity. However, this move has raised significant red flags, particularly given the NSA's history of mass surveillance and data collection without warrants. Critics, including Edward Snowden, have voiced their concerns that OpenAI's AI capabilities could be leveraged to strengthen the NSA's snooping network, further eroding individual privacy. Snowden has gone so far as to label the appointment a "willful, calculated betrayal of the rights of every person on Earth." The tech community is rightly alarmed, with many drawing parallels to dystopian fiction. The move has also raised questions about ...

  Adobe's recent update to its Terms of Service for Creative Cloud has sparked widespread concern among the creative community. The changes grant Adobe unprecedented access to users' projects, raising questions about privacy, copyright, and the very ownership of creative work. In this post, we'll delve into the key objections and legal implications of these new terms, and what they mean for professionals and individuals alike. Access to User Content: A Privacy Concern? Adobe's updated Terms of Service grant the company the right to access users' projects through automated and manual methods. While Adobe justifies this access as necessary for responding to support requests, detecting technical issues, and addressing security concerns, many users are uneasy about the potential for privacy violations and the unauthorised use of copyrighted material. This lack of transparency has led to fears that Adobe may use users' work to train its AI systems without their conse...

There have been some good attempts by organisations and individuals at developing policies on AI for third sector organisations. Some of the fundamental challenges that have been difficult to capture is the rapid evolution of AI. This rapid pace of AI development, particularly the potential for artificial general intelligence (AGI) by 2027/28 as suggested by scaling laws, poses significant challenges for charities in keeping their policies up-to-date. Here are some key considerations which may be of use to your organisation: Establish Principles and Ethical Frameworks Charities should establish clear principles and ethical frameworks to guide their use of AI, rather than relying solely on specific use cases or technical details that may quickly become outdated. These principles should align with the charity's mission, values, and commitment to beneficiaries, while addressing issues like transparency, accountability, privacy, and bias.[1][3] Adopt Agile and Iterative Policymaking Gi...

  Leopold Aschenbrenner's 'Situational Awareness, the decade ahead  Situational Awareness, the decade ahead . June 2024' may turn out to be the most significant publication on AI safety to date. Unlike a lot of theoretical musings from highly intelligent critics of AI systems this one has been written by an engineer, who was until recently employed by Open AI in the now disbanded former Super Alignment team.  It begins by discussing the rapid advancements in AI technology, particularly focusing on the progression from GPT-2 to GPT-4 models. It highlights that AI capabilities are evolving at an exponential rate, and there are predictions that by 2027, AI models could match or even surpass the work of human AI researchers and engineers. The text underscores the importance of understanding the trendlines in compute, algorithmic efficiencies, and unlocking latent capabilities for future AI development. Additionally, the document mentions the potential risks and challenge...

  Neel Nanda is involved in Mechanistic Interpretability research at DeepMind, formerly of AnthropicAI, what's fascinating about the research conducted by Nanda is he gets to peer into the Black Box to figure out how different types of AI models work. Anyone concerned with AI should understand how important this is. In this video Nanda discusses some of his findings, including 'induction heads', which turn out to have some vital properties.  Induction heads are a type of attention head that allows a language model to learn long-range dependencies in text. They do this by using a simple algorithm to complete token sequences like [A][B] ... [A] -> [B]. For example, if a model is given the sequence "The cat sat on the mat," it can use induction heads to predict that the word "mat" will be followed by the word "the". Induction heads were first discovered in 2022 by a team of researchers at OpenAI. They found that induction heads were present in ...

  'AI eliminated nearly 4,000 jobs in May' so it's reported by hallenger, Gray & Christmas, Inc. Following on from reports by IBM et al that thousands of job cuts will occur due to AI replacement, there is no need to wait for the utopia of AI allowing humans more leisure time, as that's already here, in the form of redundancies, if we are to accept the reports findings. 'With the exception of Education, Government, Industrial Manufacturing, and Utilities, every industry has seen an increase in layoffs this year.' What's particularly notable is that it's the Tech sector that's the most affected from job cuts in the US economy: 'The Technology sector announced the most cuts in May with 22,887, for a total of 136,831 this year, up 2,939% from the 4,503 cuts announced in the same period last year. The Tech sector has now announced the most cuts for the sector since 2001, when 168,395 cuts were announced for the entire year. ' Another reason ...

  This blogpost will be added to over a few days, maybe weeks as it is in response to a report that has been published today, 13th June, 2023. I am writing this in the morning, I will need time to read it through in detail. However it is important enough for me to give my initial impressions.  On first glance it seems a comprehensive report with some interesting areas for debate, acknowledgement of the potential for the transformative effect on states of such technology, yet rather predictable solutions being offered that are too state orientated. It ultimately seems to be about power. How the power of corporations co-exist with the power of the state and what a future symbiotic co-existence might look like. There are the now usual calls for the UK's state to be elevated as a centre of AI Safety (which seems geopolitically unrealistic). The potential 'benefits' seem overplayed and the potential dangers underplayed.  One fear, such interventions are beginning to bring abou...

  Home Assistants have been pushed out to wealthier populations which much glee by tech companies over the last few years. Selling the dream of dominating your home environment, by the 'master's voice', from turning on and off lights to preparing your electric car, and eventually to you home robots that will cook and clean for you, if they can ever get around to dealing with changes of floor levels. So rather than waiting for the dream to be complete and the tech companies to sell you more product, what if you could code it yourself. Well someone has tried: 'As much as we like technology, what humans love more is control and predictability. We're afraid of wild beasts with fangs, claws, and venom because we don't know how wild animals will react to us. Like the untrained, we can't risk our safety because it's difficult to protect against something that's unpredictable. Based off of some comments that I've seen in conversations around the internet...

  Claude is an LLM from Anthropic, which now has a neat trick up its sleeve, you can ingest several thousands of words into it's prompt window and ask questions of that document immediately. It's no wonder that the company advertise amongst the use cases for this model legal firms. Two Claude models were launched in May, with two different pricing structures.  I've not mentioned Anthropic before, and haven't read it's AI Safety framework as yet, which I'll have to rectify. But I not that in their paper, ' Training a Helpful and Harmless Assistant with Reinforcement Learning from Human Feedback ', they stated under limitations: We’ve pragmatically defined an aligned assistant as an AI that is18 helpful, honest, and harmless. We are optimistic that at present capability levels, the techniques we have discussed here provide a reasonable approach to achieving helpfulness and harmlessness. However, although our techniques improve model honesty, we believe we ...

  I wanted to conduct a little experiment, as a follow up to some posts which assert bias and a USA centric, hegemonic view of the world. I apologise now, that this is a necessarily long post. Please bare with me as the results that follow may raise your eyebrows and lead to some serious questions. The methodology is clear, it may not be perfect, but you can make of it what you will, and repeat it yourself with a topic of your choosing. I used GPT-4, via Perplexity AI (As it makes the sources more apparent) to suggest policy solutions to a real world problem, the economic state of the UK economy, in order to ascertain the bias in it's chosen sources and the effect this would have upon the answer(s).  I chose the field of economics as, for me, any differences in the given answers would be rapidly apparent, as I have informally studied economics since the 2008 Great Financial Crash. I'm no self-proclaimed expert but would hope I've learnt sufficient for this experiment to be ...

I have seen a few discussions and lectures from Harari on the subject of AI, this though may be the best so far. The questions are pointed, which certainly helps. Harari tends to bring a different perspective to the debate on AI safety, which is of value. It's well worth watching the whole video, below is a snippet.  Harari: So, we need to know three things about AI. First of all, AI is still just a tiny baby. We haven't seen anything yet. Real AI, deployed into the world, not in a laboratory or in science fiction, is only about 10 years old. If you look at the wonderful scenery outside, with all the plants and trees, and think about biological evolution, the evolution of life on Earth took something like 4 billion years. It took 4 billion years to reach these plants and to reach us, human beings. Now, AI is at the stage of, I don't know, amoebas. It's like 4 billion years ago, and the first living organisms are crawling out of the organic soup. ChatGPT and all these w...