๐ค A New Era of Web Usage Unfolds
Dear curious minds,
AI is set to reshape our web experience, and currently, Google isn't the one steering this revolution.
In this weekโs issue, I bring to you the following topics:
AI-Browsing Assistance: Brave Leo Now Powered by Mixtral
AI-Powered Search: Revolutionizing How We Find Information
Google's Lumiere: The Future of AI-Generated Videos?
Unlocking the Secrets to Powerful Branding: Insights from Alexandra Watkins
If nothing sparks your interest, move on, otherwise, let us dive in!
๐ฆโจ AI-Browsing Assistance: Brave Leo Now Powered by Mixtral
Brave's desktop browser update to version 1.62 introduces Mixtral 8x7B, a high-performance LLM that promises enhanced speed and accuracy in responses, as the core of Leo, their AI browser assistant. Leo was already covered in its first version in an issue from last November.
Released by Mistral AI, Mixtral 8x7B has quickly become a favorite in the developer community as the currently best performing open-source model, outshining other models like ChatGPT 3.5 and Llama 2 in performance metrics. Mixtral got initially covered in an issue from last December.
Mixtral supports multiple languages and can generate code, aligning with Brave's commitment to open-source models and community engagement.
Users can choose between different models in the free and premium plans. In my evaluation, I could use the Mixtral model six times before I was asked to go premium or switch to the free Llama 2 13B model.
The model is used in the browsing assistance named Leo, which is accessible via the browser sidebar. Besides general answers, Leo can access the content of the current page to answer your questions. One of the main use cases is the summarization of the current page.
Brave maintains strict privacy standards with Leo, using techniques like reverse proxy and immediate discarding of responses to ensure user interactions remain anonymous and secure.
My take: The integration of Mixtral 8x7B into Brave Leo marks a significant leap forward in AI-assisted browsing, blending cutting-edge technology with a commitment to user privacy. As AI becomes more embedded in our online tools, Brave's approach serves as a model for balancing innovation with the need for secure and private digital spaces.
๐๐ AI-Powered Search: Revolutionizing How We Find Information
An issue from December last year already investigated the AI web search landscape. Meanwhile, I came across two new solutions which I want to highlight in this issue.
In the mentioned coverage, Perplexity emerged as the hidden champion due to its fast execution of search request and extensive linking of sources. With around 100k followers on ๐ (formerly Twitter), Perplexity is a popular and widespread option in the AI-powered web-search space.
An AI-powered search I did not know so far is named Exa, recently renamed from Metaphor, which has around 8k followers on ๐. Their documentation states that because of the underlying transformer architecture, which in a nutshell predicts the most likely next word, users should phrase their questions as the beginning of the answer they are looking for. As an example, โWhatโs the best way to use AI for search?โ should be reformulated as โThis is the best way to use AI for search:โ. However, the current version has a feature named "autoprompted" which is a prompt optimizer that reformulates a user input as described (likely done with an LLM ๐ ).
The Browser Company which creates the Arc browser published Arc Search, also named โBrowse for Meโ. It is the most recent entrant in the AI-search space and, so far, exclusive to iOS. This solution distinguishes itself by using generative AI to construct websites from search results, offering a unique and immersive search experience that goes beyond traditional result listings. If you have 14 minutes, their keynote about the new way of browsing is worth a watch.
My take: The AI-search space is buzzing with innovation, from the quick AI-powered searches of Perplexity to the intuitive query optimization of Exa, and the groundbreaking approach of Arc Search in creating dynamic web content from search results. The latter may seem revolutionary, but fundamentally, a Google Search result is also just a webpage created for you, which you can share with others through its link. Nevertheless, as these new players redefine what's possible, it's an exciting time to watch and see how giants like Google will respond. Will they unveil a game-changing solution that leapfrogs over these advancements, or will they incorporate these innovations into their own offerings? Regardless of what happens next, I am confident that our web search experience will be completely reshaped this year.
๐ฝ๏ธ๐ค Google's Lumiere: The Future of AI-Generated Videos?
Google introduced Lumiere, an AI-based video generation tool that creates videos from text prompts and guiding images.
Lumiere's capabilities were showcased in a 2-minute-long video by Google's Research, demonstrating its ability to generate coherent, high-quality videos up to five seconds long from simple text inputs.
Examples of Lumiere's output include animating famous paintings and generating realistic animal videos, although some CGI elements like dog faces may not be perfectly realistic.
Lumiere also offers video editing tools that allow users to change the style of subjects in videos dramatically, making video editing more accessible to non-experts.
The technology behind Lumiere is described as a "space-time diffusion model," which processes videos in multiple space-time scales for more realistic and coherent motion, contrasting with traditional methods that synthesize keyframes and then fill in gaps.
Despite the impressive demos, Google has not provided details on availability or user accessibility.
My take: Lumiere's introduction is a significant leap in generative AI video creation and seems to be ahead of the competition covered in an issue from December 2023. The generations of all current AI video tools are only few seconds long. I ask myself why this limitation exists, as publicly available tools like RunwayML are used to create longer videos by using the last frame as starting point for a new generation and later concatenating the results. Furthermore, this is another release from Google which teasers powerful generations, but does not make it available to users. A sad story!
๐๐ก Unlocking the Secrets to Powerful Branding: Insights from Alexandra Watkins
Alexandra Watkins shares her expertise on branding and naming on the "Creator Science" podcast episode #168, hosted by Jay Clouse.
She founded the brand naming company Eat My Words in 2005 and published the accompanying book Hello, My Name Is Awesome: How to Create Brand Names in 2019.
Some thoughts Alexandra shared in the episode which resonated with me:
A memorable brand name leverages familiar concepts, enhancing recognition and perception, which is vital for businesses.
Effective rebranding with a strong name can transform business perceptions and outcomes, underscoring the importance of careful consideration in the naming process.
Taglines and creative domain names can elevate a brand's identity, offering a strategic workaround for unavailable domains.
The smile test lists five qualities of super sticky names:
S - Suggestive - Name should suggest something positive about your brand
M - Memorable - Name should be based on existing knowledge to be easily remembered
I - Imagery - Name should evoke a visual image
L - Legs - Name can lend itself to a theme you can extend
E - Emotional - Name should resonate emotionally with people
In contrast, the scratch test names seven dealbreakers which to avoid. Even one dealbreaker could come back to haunt you:
S - Spelling challenged - Looks like a typo
C - Copycat - Don't copy another brand
R - Restrictive - Could outgrow the name over time
A - Annoying - Causes frustration for people
T - Tame - Too boring or limiting
C - Curse of knowledge - Only you understand the meaning
H - Hard to pronounce - Easily mispronounced
My take: Hearing Alexandra Watkins talk about how to pick the best name for a brand was really exciting. She shared some smart tips, showing how a good name can make a big difference in how people see and like a brand. Alexandra's tips on choosing a name that really speaks to people made me think a lot. It showed me how important a name is and how it can really show what a brand is all about. I created a custom Brand Name Generator and Evaluator GPT which you can use as a paid ChatGPT customer to create or check a brand name with her smile and scratch test.
This publication is free. If you would like to support me, please recommend this newsletter to anyone you think would enjoy it.
Disclaimer:ย This newsletter is written with the aid of AI. I use AI as an assistant to generate and optimize the text. However, the amount of AI used varies depending on the topic and the content. I always curate and edit the text myself to ensure quality and accuracy. The opinions and views expressed in this newsletter are my own and do not necessarily reflect those of the sources or the AI models.