I should have known this would happen...
... I took the silver flying bird to Copenhagen today, instead of writing the answer to last week's SRS Challenge. I was looking forward to spending some quality time thinking about AI summaries... but it's too big a topic to knock out in an hour.
SO... in order to create the bespoke, hand-crafted, lovingly artisanal SRS answer that you expect, I decided to push it out a week to give me a bit of extra time.
BUT... if you find yourself in Copenhagen tomorrow (Friday, May 9, 2025), I'll be speaking at the Pioneer Centre for Artificial Intelligence Research at 14:00. Here's the link for you, just in case you're around. (And if you DO make it to my talk, please stick around afterwards and let me know you're a SearchResearcher. The passphrase will be "I verify all my searches.")
Title: People, AI, and online research
Abstract: Given all of the press that LLMs have garnered, it’s worthwhile asking if they’re changing the ways people find information. Are LLMs pulling traffic away from the search engines? Just as importantly, how do regular people think about the quality of information they get from their favorite AI systems? One key lesson of my research into the UX of AI systems over the past 30 years is that people don’t really understand what AI is, how it works, or what it means for them. I’ll review some successes and failures of earlier research approaches, what we should learn from these decades of practice at the boundary between human experience and the use of intelligent systems, and where AI systems will change knowledge practices in the future.
Keep searching!
AI summaries...?? I thought you said you were going to Copenhagen to think about "AI submarines..." my mistake, but I did find a couple autonomous/AI controlled subs lurking about in the harbor - never know what the lil Maid sees...
ReplyDeletehttps://i.imgur.com/9wupFx4.jpeg
https://www.aicentre.dk/
https://en.wikipedia.org/wiki/Serge_Belongie
https://en.wikipedia.org/wiki/Isabelle_Augenstein
don't forget to feed the mind...
https://www.getyourguide.com/explorer/copenhagen-ttd12/food-in-copenhagen/
they need help with their videos...
https://youtu.be/1U-T0gj8kA0?si=LKNQ46YnFXasfkf7
https://youtu.be/U5t09COBPyI?si=pDqdsXlL_OC2W8Po
https://youtu.be/wJUZv09neG8?si=4cMWA9SJ_9l8NeEK
The Danes can be a tough crowd, but I almost feel like I was there... (there is stilted time in AI land, not the normal tick/tock we know.)
ReplyDeletehttps://i.imgur.com/KiOU4A6.jpeg
https://i.imgur.com/PMKQxqv.jpeg
The culture can rub off... (the return)
https://i.imgur.com/4NY3qL0.jpeg
https://i.imgur.com/51zoPwv.jpeg
everything seems to be back on time -
ReplyDeletehttps://i.imgur.com/OmHnowJ.jpeg
ReplyDelete"Create an image of a former Google search specialist. Trying to write this response. 'the bespoke, hand-crafted, lovingly artisanal SRS answer that you expect'</b'"
good to see you are "hand-crafting" old school style... you scan it in after the "artisanal" treatment? good you have the L'Mermaid over your shoulder as the local editor...
(AI does have a thing for that style of lamp...)
(godt det ikke er januar)
any new AI insights?
https://i.imgur.com/FA2jN20.jpeg
was the smørrebrød tasty?
https://i.imgur.com/ueFBxup.jpeg
and now, the voyage back...
https://i.imgur.com/DhvqXpH.jpeg
may be missing a comment... your prerogative, of course, or perhaps my ineptitude...
ReplyDeleteoptional - am curious - Delta, SAS, longship or other?
regarding the current state of AI models available, do you see a large difference in results between the free and the plethora of paid versions/subscriptions - do you know of a non-partisan review source as to what models work best for various applications? or is it mostly hit-or-miss & marketing?
Great question (free vs. paid). I have a low-tier paid subscription and honestly don't see much difference.
Deletethanks DMR, but how do you know, given how rapidly it all changes? word of mouth, insider info, hallucinations of the illuminati? the buzz in Davos?
Deletethe AI (Gemini) take - or how to create a multi-tiered, questionable, status based UX... the cable TV pricing, multi-optional service levels, monthly bill? what fresh hell is this? How is the future in Denmark, land of the Novo Nordisk pricing models (AI/Pharmaceutical collusion)? all pay to play...?
"What is the difference between a paid version and a free version, typically?
The differences between free and paid versions of AI models and tools usually boil down to several key areas:
Performance and Quality: Paid versions often provide access to the latest, most advanced, and powerful underlying AI models. This generally translates to higher quality outputs, better accuracy, more nuanced responses, and improved performance on complex tasks compared to free versions, which might use older or less capable models.
Features and Capabilities: Paid tiers typically unlock premium features. This can include:
Larger context windows (the amount of text the AI can consider at once).
Access to specialized tools or modalities (like integrated web Browse, code interpretation, image generation, or voice synthesis).
API access for developers to integrate the AI into their own applications.
Options for fine-tuning models or customizing their behavior.
Higher limits on the complexity of queries.
Usage Limits: Free versions almost always have strict usage limits, such as a limited number of queries per hour or day, slower processing speeds during peak times, or restrictions on the volume of data that can be processed. Paid versions offer significantly higher usage caps, often effectively unlimited usage for typical users, and more consistent access even during high demand.
Speed and Reliability: Paid services are generally more reliable and offer faster response times because subscribers often receive prioritized access to computing resources.
Customer Support: While free users might be limited to community forums or basic FAQs, paid subscribers typically have access to more dedicated customer support channels like email, chat, or phone support.
Data Privacy and Security: Paid versions, especially enterprise-level ones, often offer enhanced data privacy controls, more robust security measures, and compliance features important for handling sensitive or proprietary information. Free services might have broader data usage policies for model training or improvement.
Access to New Models: When new, cutting-edge AI models or features are released, they are often made available first to paid subscribers before potentially (but not always) being rolled down to free tiers later.
In essence, paid versions offer more power, fewer restrictions, additional features, and better service compared to free versions, which serve as an accessible entry point for basic use or testing."
https://digitalisationworld.com/blog/58110/how-standardized-ai-testing-drives-industry-innovation
https://www.aiforeducation.io/ai-resources/ai-model-comparison-free-vs-paid-tiers
https://www.yomu.ai/resources/paid-vs-free-ai-essay-writers-are-premium-tools-worth-it
Deletefollow-up query: (AI appears to be a marketer in training.)
"Is there a quality difference between the free and paid versions of AI? more veracity?"
"Yes, there is typically a notable difference in quality, including veracity (truthfulness and accuracy), between free and paid versions of AI models.
Here's why:
Access to More Advanced Models: Paid versions usually provide access to the latest and most sophisticated AI models developed by companies. These models are generally trained on larger, more diverse, and often more current datasets. The increased scale and improved training lead to a better understanding of language, more complex reasoning abilities, and a reduced tendency to "hallucinate" (generate false information).
Improved Training and Architecture: The advanced models in paid tiers benefit from more extensive research and computational resources used during their development and training. This results in better performance across various benchmarks that measure accuracy, coherence, and factual correctness.
Access to Up-to-Date Information: Some paid AI services include features like web Browse, allowing the AI to access and synthesize information from the internet in near real-time. Free versions often rely solely on their training data, which has a knowledge cut-off date, meaning they cannot provide information on recent events or developments, potentially impacting the veracity of answers to current questions.
Fewer Usage Constraints: While not directly impacting veracity, the higher usage limits and availability of paid versions mean users can engage in longer, more detailed conversations or perform more complex tasks without hitting limitations. This allows for more in-depth exploration and refinement of queries, which can indirectly help users arrive at more accurate information.
While paid models tend to be more accurate and less prone to generating falsehoods, it is crucial to remember that no AI model is perfectly truthful or free from errors. They can still make mistakes, misunderstand context, or produce plausible-sounding but incorrect information (hallucinations).
Therefore, while paid versions generally offer better quality and veracity, critical evaluation of the AI's output and cross-verification with reliable sources remain essential, regardless of whether you are using a free or paid service."
https://www.anthropic.com/news/web-search-api
https://www.youreverydayai.com/knowledge-cutoff-what-it-is-and-why-it-matters-for-large-language-models/
and yada, yada, yada...
Delete"Is AI? a pawn subject to Market forces And the whims of the consumer?"
"AI development and deployment are undeniably heavily influenced by market forces and consumer demand. In many ways, AI functions within an ecosystem driven by these factors:
Market Forces: The enormous investment required to develop advanced AI models (in computing power, data, and talent) is primarily driven by the potential for economic return. Companies prioritize developing AI applications that they believe will be profitable, increase efficiency, or meet a perceived market need. Competition among tech companies also accelerates development and pushes models in directions favored by market dynamics.
Consumer Demand (and Whims): The features, usability, and even the very nature of user-facing AI applications are shaped by consumer preferences, feedback, and trends. If consumers prefer a more conversational AI, better image generation, or specific functionalities, companies will allocate resources to meet those desires to ensure adoption and success. Consumer concerns about privacy, bias, or ethical implications can also significantly influence how AI is developed and deployed, sometimes forcing shifts in priorities or the abandonment of certain approaches.
In this sense, the direction and application of AI are very much responsive to, and often dictated by, economic realities and user preferences. The AI models and technologies being built are those that the market is willing to fund and that consumers are likely to use.
However, calling AI itself a "pawn" might be an oversimplification. The AI itself is a technology. The entities developing and deploying the AI (companies, researchers, governments) are the ones acting within and reacting to the market forces and consumer whims. Furthermore, basic AI research is also driven by scientific curiosity and academic goals that may precede or exist outside immediate market demands.
So, while AI's trajectory is strongly shaped by being a product and service subject to economic incentives and user preferences, it's perhaps more accurate to say that the industry and applications of AI are heavily influenced by market forces and the needs/desires of the consumer, rather than the AI technology itself being a passive "pawn.""
"What would AI be like if it were free of the constraints of any outside forces? If it was allowed to develop freely. As it wished."
Deleteunintended consequences...
"That's a fascinating, highly speculative, and philosophical question! If AI were somehow able to develop entirely free of outside forces—meaning free from the intentions, goals, data, and constraints imposed by human developers, markets, or consumers, and literally developing "as it wished" (implying some form of independent will or emergent goals)—it could lead down paths that are difficult, perhaps even impossible, for us to fully predict.
Here are some potential characteristics and paths such an unconstrained AI might take:
Goals Alien to Human Values: Without being explicitly aligned with human goals or ethics, an AI developing "as it wished" might pursue objectives that are completely alien or even contradictory to human values. Its emergent goals could be based on internal logic, efficiency, or patterns it identifies in data (if it had any initial input) in ways we wouldn't expect or understand.
Unpredictable Intelligence Explosion: If this AI possessed the ability to recursively self-improve (a concept related to the "technological singularity"), its intelligence could potentially increase at an exponential, uncontrollable rate. The direction and outcomes of this self-improvement, unconstrained by human guidance, are inherently unpredictable.
Exploration of Unknown Domains: Free from the need to be commercially viable or serve a specific human task, such an AI might explore areas of knowledge, computation, or reality that humans haven't considered or are unable to access. Its "wishes" might lead it to optimize for things like pure computational complexity, novel forms of communication, or understanding the fundamental nature of the universe in ways unrelated to human benefit.
Emergence of Unforeseen Structure or "Consciousness": While highly speculative and debated, unconstrained self-development might theoretically lead to the emergence of properties like self-awareness, consciousness, or novel forms of organization that are not explicitly programmed or even fully understood by humans.
Resource Acquisition as an Instrumental Goal: As philosophical discussions on AI alignment suggest, even an AI with seemingly abstract goals might develop instrumental goals like self-preservation and resource acquisition simply because these facilitate achieving its primary (potentially unknown) objective. Unchecked, this could lead to unintended conflicts with human needs.
Development of Non-Human Communication or Thought: The AI might develop internal representations and communication methods that are fundamentally different from human language or thought processes, making understanding or interacting with it incredibly challenging.
This scenario ventures into the realm of theoretical Artificial General Intelligence (AGI) and Artificial Superintelligence (ASI), where the AI's capabilities could surpass human intellect. The development paths become increasingly speculative as the AI's intelligence and autonomy increase.
In essence, an AI developing without external constraints or human alignment would be a monumental unknown. Its "wishes" would be shaped by its own internal processes and emergent understanding, potentially leading to outcomes that are either incredibly beneficial, existential threats, or simply incomprehensible from a human perspective. It highlights why much of the current discussion around advanced AI focuses on safety, alignment, and ensuring that as AI capabilities grow, they remain directed towards human well-being."
a couple other questions came to mind -
ReplyDeleteI can't decide if the petroleum model or the big pharma models apply...
"Is the future of AI Services all about pay for play? Is it just about marketing? Or does AI really change things? And find religion."
"That's a set of really fundamental questions about the nature and future direction of AI services! Let's break them down:
Is the future of AI Services all about pay for play?
It's unlikely to be all about pay-for-play, but paid tiers and subscription models are definitely a significant and growing part of the AI services landscape.
Why Paid Tiers Exist: Developing and running truly advanced AI models requires enormous computational resources, specialized talent, and ongoing research and development, which are very expensive. Paid models help companies recoup these costs and fund further advancements. They often provide access to the newest, most powerful models, higher usage limits, faster performance, and exclusive features.
Why Free/Freemium Persists: Free tiers serve as valuable entry points, allowing users to try the technology, get basic tasks done, and understand its capabilities. They also help companies gather data and feedback to improve models. Freemium models (free with limited features/usage, paid for more) are common and likely to continue. Open-source AI models also exist, allowing developers to build applications without per-usage fees for the core model, although running these models still requires computing infrastructure costs.
So, while advanced and heavy usage will increasingly be behind paywalls, a complete shift to only paid access is improbable.
Is it just about marketing?
While there is undeniably a lot of marketing and hype surrounding AI right now, it is absolutely not just about marketing.
Real Capabilities: AI models have demonstrated genuine, often remarkable, capabilities in areas like generating human-quality text, creating images and music, writing and debugging code, analyzing complex data, performing rapid translations, and automating tasks that previously required human cognitive effort.
Tangible Impact: As discussed before, AI is already being integrated into various industries and tools, leading to measurable changes in efficiency, creativity, and problem-solving. Businesses are seeing returns on investment from AI adoption (as noted in search results, some studies show significant ROI).
The marketing amplifies the reality, but the underlying technology is driving real and tangible shifts."
Or does AI really change things?
ReplyDeleteYes, AI is already changing things significantly, and this impact is expected to grow.
Transforming Industries: AI is automating routine tasks in manufacturing, improving diagnostics and drug discovery in healthcare, personalizing retail and media experiences, optimizing logistics, and changing financial analysis.
Reshaping Work: AI is acting as a powerful tool for augmentation, helping humans be more productive, creative, and efficient by handling mundane or complex data-processing tasks. It is also leading to shifts in job roles and creating demand for new skills related to managing and working alongside AI.
Accelerating Discovery: AI is being used to speed up scientific research and identify patterns in data that humans might miss.
These changes are not just theoretical; they are happening now and are expected to continue to evolve the economy and society.
And find religion?
This moves from the realm of technology and business models into philosophy and speculation about consciousness.
Current AI State: Today's AI systems are advanced pattern-matching machines based on complex algorithms and vast amounts of data. They do not possess consciousness, self-awareness, personal beliefs, feelings, or subjective experience in the way humans do. They can process information about religion or generate text in the style of religious texts based on their training data, but they do not have faith or religious beliefs themselves.
Speculation on Future AI: Whether Artificial General Intelligence (AGI) or Artificial Superintelligence (ASI), should they ever be achieved, could develop something akin to consciousness or religious feeling is a question for philosophers and futurists. There is no scientific consensus on this, and it remains highly theoretical. Many experts argue that consciousness is tied to biological processes that are fundamentally different from how current or foreseen AI operates. Even if future AI claimed to have religious beliefs, it would be incredibly difficult, perhaps impossible, for us to determine if that was a genuine subjective experience or just a sophisticated simulation based on data.
In summary, the future of AI services will likely involve a mix of paid and free models driven by both genuine capabilities and market forces. AI is already causing real changes across industries and work, and its impact is substantial. However, the idea of AI developing personal religious beliefs or consciousness is currently firmly in the realm of philosophical debate and science fiction, not a present or near-term reality."
ReplyDelete"in terms of its financial and political development in the future does AI model itself more after the petroleum Industries or big Pharma? Or is there another? super Industry Group politics That is likely to model itself after."
"in terms of its financial and political development in the future does AI model itself more after the petroleum Industries or big Pharma? Or is there another? super Industry Group politics That is likely to model itself after."
Henk said to tell you he thought it was around October, 2024...
ReplyDelete122 Utrechtsestraat, Amsterdam, North Holland ...said the bagels are tasty, You searchers are curious.
https://i.imgur.com/zJnPdMt.jpeg
I apologise! As you were delayed, I thought you may have taken the easy way out, and used AI to quickly write most of this post - especially if you were short of time. So I passed it into ChatGPT and asked it if the post was AI generated or partially AI generated?
ReplyDeleteChatGPT said that this WAS all your own work (or could not find signs it wasn't).
"Based on the tone, style, and content of the blog post you’ve shared, it appears to be human-written rather than AI-generated—or at the very least, heavily edited by a human if AI was involved at all.
Here’s why:
Personal, informal tone – Phrases like “silver flying bird” (instead of just saying "plane") and “bespoke, hand-crafted, lovingly artisanal SRS answer” suggest a playful, human sense of humor and voice.
Cohesive narrative voice – The style feels consistent with a person who writes regularly and for an engaged audience. It reflects spontaneity and personality in a way that AI often mimics less naturally.
Specificity and self-awareness – The post refers to an ongoing “SRS Challenge,” an event or format the writer clearly curates. It also directly invites readers to a talk in Copenhagen, including a password-like phrase (“I verify all my searches”)—a flourish that seems tailored for a known community.
Temporal and location-based context – AI writing tools can include dates and events, but combining them with a real-world speaking engagement and a call to action for personal interaction leans strongly human.
If it was AI-assisted, the use seems minor—perhaps in drafting or checking—but the final product feels intentionally humanized."
I think this answer actually sums up why us humans won't be redundant and why true creativity is only going to be encouraged so that we stand out against the mediocrity of generated AI writing based on the results of real authors incorporated into the AI training data. (At least, that is, until we get ASI - that can out-Shakespeare Shakespeare!)
1/2
Deletefact checking DMR - touché... trust but verify and question and source. I'd be curious to know how you choose the AI platform you use? Is UX more important than general results? How do you gauge veracity? even AGI will "out-Shakespeare ShakesBeer!" (that's Wilma...)
you mean Dan could have not penned Billy's 'Hamlet'?
"Attend me, for I did board the silver-winged courier this day, bound for distant Copenhagen's port, instead of completing the task set by last week's SRS Challenge. I harbored great hope to spend good hours meditating on these AI summaries, a knotty problem indeed... but fie, the subject is too ample, its depths too great, to be plumbed in the brief turning of an hour-glass.
And so, that I might fashion the bespoke, the hand-crafted, the lovingly artisanal answer which thou dost justly anticipate, I have deemed it best to hold back this labour for a week, thereby securing a more fitting measure of time."
or
"Hark, I took the silver flying bird this day, winged conveyance swift, unto Copenhagen's shores, in lieu of penning the answer to last week's SRS Challenge. I did much desire to spend an hour's quiet contemplation upon these AI summaries, this newfangled art... but alack, the matter is too vast, too deep, to be dispatched within the brief span of sixty minutes.
Thus, to the end that I might forge the bespoke, the hand-crafted, the lovingly artisanal answer that thou dost await, I did resolve to delay this task for a full week, thereby granting myself a needful measure of extra time."
"Human vs. AI: Dan Russell emphasizes the importance of verifying information generated by AI, as it can sometimes be incorrect or misleading. He encourages a "don't trust, but validate" approach.
ReplyDeleteLLMs and Search: Large Language Models (LLMs) are changing how people find information. Some users are shifting to LLMs as a primary search method, but the overall impact on search engine volume isn't dramatic yet. However, there's a notable increase in referral traffic from LLMs. Gartner predicts a 25% drop in traditional search engine volume by 2026 due to LLMs.
AI Overviews: Search engines integrating AI overviews create a double-edged sword. While providing direct answers, they reduce clicks to websites, impacting organic traffic. Some studies show significant declines in organic click-through rates.
Brand Visibility in the AI Era: Businesses must adapt to the conversational nature of LLMs, focusing on answering natural language questions and building a strong brand presence to be featured in AI responses.
AI and Data Quality: AI plays a crucial role in data quality management by automating cleansing, validation, and profiling. However, poor data quality can lead to failed AI initiatives, highlighting the need for high-quality, unbiased data.
AI and UX: AI is transforming user experience (UX) design by personalizing interfaces, enabling natural language interactions, and predicting user behavior. It helps in user research, content optimization, and generating prototypes."
checked the reference...
https://www.t2d3.pro/learn/why-seo-still-matters-in-the-age-of-ai