Adel Zaalouk

Generative AI and LLMs can offer many benefits, but come with a catch: an increase in interdependencies that exacerbates today's hidden ML tech-debt thetechnomist.com/p/generati…

I recently revisited a fascinating interview with Jeff Bezos from 1999, and it's packed with timeless insights on building customer-centric products. Bezos highlights that true customer obsession goes beyond just offering good service, it's about anticipating needs and leveraging technology like AI as a tool (not as the main product) to create truly exceptional experiences. Some takeaways: 👉 Customer obsession is the foundation: Bezos believes that prioritizing the customer above all else is the key to long-term success. 👉 Technology as an enabler: He views technologies like the internet and AI as tools to enhance the customer experience, not as ends in themselves. 👉 Proactive problem-solving: Bezos encourages anticipating customer needs and proactively inventing solutions before they even realize they have a problem. What are your thoughts on Bezos' customer-centric approach? Full post: thetechnomist.com/p/customer……I recently revisited an interview with Jeff Bezos from 1999, and it's packed with timeless insights on building customer-centric products. Bezos highlights that true customer obsession goes beyond just offering a good service. It's about anticipating needs and leveraging technology like AI as a tool (not as the end product) to create truly exceptional experiences. Some takeaways:

It’s surprising to me that in 𝟭𝟵𝟲𝟰, we had the same interfaces we have now (chatting with intelligence). I’d say Eliza is probably the closest to ChatGPT conceptually, but the tech, the intelligence itself, was rule-based. It was creative but still followed a deterministic logic. That said, it still beat GPT 3.5 in the Turing test (source: arxiv.org/abs/2310.20216)! Here is Eliza's demo: youtube.com/watch?v=RMK9Aphf… Want to chat with Eliza? Here you go (It can be a tad philosophical!): psych.fullerton.edu/mbirnbau…… Based on the Eliza experiment, it turns out it's not that hard to trick people into thinking that real intelligence is behind the screen when, in fact, it was just a set of rules: "𝘑𝘰𝘴𝘦𝘱𝘩 𝘞𝘦𝘪𝘻𝘦𝘯𝘣𝘢𝘶𝘮 𝘸𝘳𝘰𝘵𝘦 𝘵𝘩𝘦 𝘱𝘳𝘰𝘨𝘳𝘢𝘮 𝘌𝘓𝘐𝘡𝘈 – 𝘦𝘴𝘴𝘦𝘯𝘵𝘪𝘢𝘭𝘭𝘺 𝘢 𝘙𝘰𝘨𝘦𝘳𝘪𝘢𝘯 𝘵𝘩𝘦𝘳𝘢𝘱𝘪𝘴𝘵 𝘵𝘩𝘢𝘵 𝘩𝘢𝘴 𝘢 𝘯𝘢𝘳𝘳𝘰𝘸 𝘧𝘰𝘤𝘶𝘴 𝘰𝘯 𝘴𝘦𝘹 𝘢𝘯𝘥 𝘳𝘦𝘵𝘶𝘳𝘯𝘴 𝘪𝘯𝘱𝘶𝘵 𝘢𝘴 𝘲𝘶𝘦𝘴𝘵𝘪𝘰𝘯𝘴 – 𝘪𝘯 1966, 𝘮𝘦𝘳𝘦𝘭𝘺 𝘢𝘴 𝘢 𝘤𝘭𝘦𝘷𝘦𝘳 𝘱𝘳𝘰𝘨𝘳𝘢𝘮 𝘧𝘰𝘳 𝘦𝘯𝘵𝘦𝘳𝘵𝘢𝘪𝘯𝘮𝘦𝘯𝘵.[1] 𝙄𝙩 𝙖𝙘𝙘𝙚𝙥𝙩𝙚𝙙 𝙩𝙮𝙥𝙚𝙙 𝙞𝙣𝙥𝙪𝙩 𝙖𝙣𝙙 𝙧𝙚𝙨𝙥𝙤𝙣𝙙𝙚𝙙, 𝙤𝙛𝙩𝙚𝙣 𝙖𝙥𝙥𝙧𝙤𝙥𝙧𝙞𝙖𝙩𝙚𝙡𝙮. 𝘛𝘩𝘦 𝘱𝘳𝘰𝘨𝘳𝘢𝘮 𝘸𝘢𝘴 𝘰𝘧𝘵𝘦𝘯 𝘣𝘦𝘭𝘪𝘦𝘷𝘢𝘣𝘭𝘦 𝘢𝘴 𝘢𝘯 𝘪𝘯𝘵𝘦𝘭𝘭𝘪𝘨𝘦𝘯𝘵 𝘣𝘦𝘪𝘯𝘨, 𝙬𝙝𝙞𝙘𝙝 𝙬𝙖𝙨 𝙥𝙖𝙧𝙩𝙞𝙘𝙪𝙡𝙖𝙧𝙡𝙮 𝙨𝙪𝙧𝙥𝙧𝙞𝙨𝙞𝙣𝙜 𝙩𝙤 𝙒𝙚𝙞𝙯𝙚𝙣𝙗𝙖𝙪𝙢, 𝘸𝘩𝘰 𝘯𝘦𝘷𝘦𝘳 𝘪𝘮𝘢𝘨𝘪𝘯𝘦𝘥 𝘵𝘩𝘢𝘵 𝘪𝘵 𝘸𝘰𝘶𝘭𝘥 𝘣𝘦 𝘤𝘰𝘯𝘷𝘪𝘯𝘤𝘪𝘯𝘨 𝘢𝘵 𝘢𝘭𝘭". ~ from web.stanford.edu/class/sts12…… As intelligent as we are being human, we are easy to trick most of the time. This begs the question, are we being tricked by the LLM's ability to predict the next token from context these days (and are we evaluating for how good an LLM can trick humans)? Do we really think there is reasoning going on? More on Eliza and similar AI from history in my latest post: thetechnomist.com/p/history-…

Visualized using the data from Médéric HURIER's post on LinkedIn: lnkd.in/eB_e-Vu2 Looks like Gemini Flash's pricing is impressive! Even if it means they are loss-leading for a while, they can make up for it being Google! Would be interested how it fares in benchmarks against Opus, and GPT-4o. Regardless, exciting times  and great news for all consumers around the world. More competition please :) #AI #Pricing