Banks will charge you when you spend or transfer money abroad. That’s why over four million people have switched to Revolut. Sign up and get Revolut Premium free for 3 months!
Friends,
See my what to read section scroll below. But if you are here for the AI, keep reading.
I wasn't expecting Gemini (watch and read this whole page), but Google's latest language model, is a powerful AI model that is positioned as a competitor to OpenAI's GPT models. Google claims that Gemini Pro outperforms GPT-3.5 in several industry benchmarks, and Gemini Ultra even surpasses GPT-4 in some cases. Does it matter? Not really - because it shows there are multiple horses in this race.
The marketing by Google was genius. Slickly created and while there was some concern about the promotional video, Google was criticized for featuring staged and edited outputs, which raised questions about how accurately the model's capabilities were represented. Again, the marketers did a great job, because if they're not quite there, they will be there soon - and no doubt we will hear from OpenAI, Meta, Anthropic etc as each tries to supersede the other. The fact that Google's models are closing the performance gap with OpenAI's models indicates a healthy level of competition in the field of AI research and development will ultimately benefit users and researchers by driving innovation and pushing the boundaries of what AI can do - hopefully in a responsible way.
Gemini offers unique features, including the ability to perform scientific research, code, and provide a user-friendly interface. It is integrated into various Google products and has access to extensive data resources, giving it a strategic advantage. In the competitive chatbot landscape, ChatGPT is currently the dominant player due to its data advantages and strong performance. Winner-take-all dynamics may come into play, with users gravitating towards the best chatbot available - but I'm not quite sure it's just a game of chatbots.
The concept of a multibot ecosystem, where users interact with various chatbots within a single interface, is considered a possibility. And I think we will see the Customization of chatbots for specific purposes as a direction in chatbot development. That said, a single dominant destination chatbot, possibly ChatGPT, could emerge as the leader based on its current position and data advantages - in the same way that Google won out at Search Engines in a crowded field.
Google's Gemini AI model appears to be a significant advancement in the field of natural language processing and generative AI. Google claims that Gemini Pro outperformed OpenAI's GPT-3.5 in six out of eight industry benchmarks, and Gemini Ultra performed even better, beating the newer GPT-4 in seven of the eight benchmarks. This suggests that Google's Gemini models are competitive and, in some cases, superior to OpenAI's GPT models. What I think is important in understanding comparisons is that the effectiveness of AI models can vary depending on the specific tasks and datasets they are evaluated on, so any comparison needs to be taken in context. In some cases it is marketing, in others it's large data sets, in others it's efficiency.
As all of this continues to evolve, both Google and OpenAI will likely continue to refine and develop their models, leading to further advancements in generative AI.
I think that gives Google Gemini a step up is its ability to perform scientific research, code, and offer a user-friendly interface. If you watch the videos, these features suggest that Gemini has unique strengths that could make it competitive in the chatbot market. Gemini's integration into various Google products, like Chrome, Gmail, and Docs, gives it a strategic advantage by reaching a wide user base - ie Google already has this foothold. Additionally, its access to Google's vast data resources could enhance its performance and make it more appealing to users. If this turns into a winner-take-all type situation, ChatGPT is in the lead, but Google with Gemini could come racing through. However, the idea of users interacting with multiple chatbots within a single destination chatbot interface is an intriguing concept. This aligns with the idea that users maintain numerous social relationships, and the same could apply to chatbots, each offering different expertise or personalities. We will have to see what happens with customization, bot by bot. The ability to iterate and customize GPT models to cater to specific needs or situations is seen as a potential direction in chatbot development, allowing users to interact with different chatbots for various purposes.
While Gemini was impressive in many ways, the tech community reacted negatively to the exaggerated marketing of Gemini. However, even with these slight oversteps, Gemini remains an impressive AI product with promising capabilities. Its integration with Google Cloud and access to Google's data resources position it well. I think you can gauge sentiment with what happens to the share price. Up by 5% adding approximately $80 billion to the company's market cap. Google's previous attempt with Bard, which faced criticism and market consequences due to a rushed launch, signalled the urgency of responding to the threat posed by ChatGPT. Now, Google has positioned Gemini as a multimodal AI model trained on images, video, audio, and text, setting it apart from previous AI models. While ChatGPT had introduced image processing capabilities, Google's emphasis on Gemini's multimodal nature suggests an attempt to differentiate itself in the AI market. And I loved the way that it is tiering its proposition strategically linked to physical products (phones laptops) and actual needs.
This is what an AI Arms race looks like. Marketing and jockeying for position is happening at pace. (oh, and I can see a glimmer of industries like education being ripped apart)
What to read
Last week I talked about reading. This week is a curation of books to read from the people who know.
Time’s 100 must-read books of 2023
Lithub’s 38 Best Books of 2023
The Conversation’s Best Books of 2023
55 Books Scientific American Recommends in 2023
Esquire’s 20 Best Books of 2023
The 10 best books of the year, according to BookTok
Stay Curious - and don’t forget to be amazing,
Key ideas from bestselling non-fiction books, distilled by experts into bitesize text and audio. Explore our vast library of over 5,500 titles and stay up-to-date with 40 new titles that are added each month. Join the reading revolution!
Here are my recommendations for this week:
Now
Skillshare is a platform that offers thousands of classes in a variety of subjects. From creative arts to business and technology, there's something for everyone. Expert instructors guide students every step of the way and a community of passionate learners provides support and inspiration. Get One Free Month of Skillshare Explore thousands of hands-on creative classes with one inspiring membership.
Person of the Year: Taylor Swift - surprised it wasn’t Sam Altman. But he’s CEO of the Year (although not for a few days last month)
7 Therapists Share Their Best Advice for Coping With Existential Dread and Grief
The world is a mess right now, but there are ways to cope.
The PISA rankings 2022 - The pisa ranking came out this week, showing how well kids do in different countries, specifically in Maths, Reading & Science. But the report is also insightful in seeing what happened to kids post Covid. There’s also a pertinent quote about use of digital at this early age (11): “Digital distraction has a strong association with learning outcomes. Students who reported being distracted by other students using digital devices in some, most or every maths class scored 15 points lower in PISA maths tests than those who barely experienced this. This represents the equivalent of three-quarters of a year’s worth of education, even after accounting for students’ and schools’ socio-economic profile. The amount of time spent on digital devices at school also seems to have an effect. While learning outcomes were often better for students who used digital devices for learning between one to five hours a day than for those who never used them, students who used them more than an hour a day for leisure – social media apps, browsing the internet or games – saw a big drop in maths scores."
Earth on verge of five catastrophic climate tipping points, scientists warn
Humanity faces ‘devastating domino effects’ including mass displacement and financial ruin as planet warms
Are Low-Fat Dairy Products Really Healthier? For decades, experts have said that less is more when it comes to dairy fat and health. But recent research has called this into question.
Next
Chat with your enterprise databases using secure generative AI and empower business users in your team to do their own data analyses in seconds. Check out Hal9
Freeing Ourselves From The Clutches Of Big Tech - Right to repair and other efforts to liberate technology from monopolistic corporations is a precondition for winning many vital societal battles.
Why 2023 was the year of the e-bike and not the self-driving car: E-bikes and self-driving cars both entered the year with momentum. Only one maintained it.
The real research behind the wild rumors about OpenAI’s Q* project - OpenAI hasn't said what Q* is, but it has revealed plenty of clues.
AI and Mass Spying: Spying is the business model of the internet and with AI, you can put together a different puzzle
The race to 5G is over — now it’s time to pay the bill - Networks spent years telling us that 5G would change everything. But the flashiest use cases are nowhere to be found — and the race to deploy the tech was costly in more ways than one.
Free your newsletters from the inbox: Meco is a distraction-free space for reading newsletters outside the inbox. The app has features designed to supercharge your learnings from your favourite writers. Become a more productive reader and cut out the noise with Meco - try the app today
If you enjoyed this edition of Box of Amazing, please share and help me grow this group. If you share it on LinkedIn, please tag me so that I can thank you.