🥊 OpenAI’s SearchGPT vs Google

ALSO: Mistral reveals AI model Large 2

Estimated Read Time: 4 minutes

Search engines might be in serious trouble. And so are websites which rely on them. More and more AI tools are searching the web for you and distilling information down into their basest form. And as this happens, websites will lose actual visitors and go out of business. Now OpenAI has introduced their search engine alternative, SearchGPT.

• 🤯🥊 SearchGPT is OpenAI’s answer to Google Search.

• 🫢 Researchers removed Llama 3’s safeguards in 3 minutes.

• ⚡️ Google Gemini ramps up their free tier.

• 🔥 Mistral AI reveals newest model.

🤯🥊 SearchGPT is OpenAI’s answer to Google Search

What happened: 🔫 Shots fired. OpenAI unveiled SearchGPT, a new search feature designed to provide timely answers from web sources. It’s a direct competitor to Google Search and a replacement for all search engines.

The details: 

• SearchGPT will give “timely” answers to questions, drawing from web sources. Much like Google’s Gemini search answers. (Yay more scraping…)

• The UI will be similar to ChatGPT. Users type queries and receive information, photos, and relevant web links. There’s also a sidebar for follow-up questions and related searches.

• SearchGPT will be able to use your location to give you local information (local restaurants, weather).

• In the support doc, OpenAI mentions SearchGPT gathers location info from third party sources unless you toggle a setting to allow SearchGPT to access your location directly.

• OpenAI claims sites can be surfaced in search results even if they opt out of AI training… but I don’t believe them.

• It’s currently available for a select group of early testers. You can join the waitlist here.

Why it matters: Google’s been the king of search for a long time. While Perplexity is a GREAT tool, it doesn’t replace Google for most people. But this has that potential. Which is why Google’s been in such a panic to integrate Gemini into their search results; they’re scared.

This also poses a second problem: AI eating its own tail. If everyone uses AI to scrape answers from the web… then how will websites fund themselves? (Aside from big publishers who can make monetized deals.) And what if they have false or misleading info? AI is basically creating a walled-garden for the internet.

🫢 Researchers removed Llama 3’s safeguards in 3 minutes. In a recent paper, researchers found a way to bypass safety protocols with ease. The study evaluates three fine-tuning methods and demonstrates that an attacker with model weights can strip safety features from Llama 3 models quickly—within one minute for an 8B model and 30 minutes for a 70B model using a single GPU. Wow. This is why cybersecurity is such a safe job right now.

📱 Meta launches AI-generated selfies feature. Meta AI’s assistant, Imagine Yourself, now supports more languages and can create stylized selfies. It’s available in 22 countries, and now supports languages like French, German, Hindi, Italian, Portuguese, and Spanish.

🎨 Adobe released new Firefly tools for Photoshop and Illustrator. Now you can create AI-generated textures and images from prompts. These are the “Generative Shape Fill” feature for Illustrator and a text-to-image generator for Photoshop. Adobe offers these tools with limited free generative credits as part of Creative Cloud plans. Illustrator also added a Dimension tool for measuring and a Mockup tool for realistic logo placement. Despite not charging extra for generative AI, Adobe has already seen customers upgrade to higher plans for more credits. I’m not surprised. Firefly’s features are among the most useful in generative AI. I highly recommend using them.

🤖 Former Tesla robot team-head launches new robotics startup, Mytra, a robotics startup focusing on warehouse automation. You know, the most sought-after use for robots. Their plan is to differentiate themselves with the ability to manage heavy payloads and dynamic operations. So not just light little boxes. The company has secured a pilot with Albertsons and has garnered interest from several Fortune 50 companies. Recently, Mytra closed a $50 million Series B funding round, bringing its total funding to 💰 $78 million, with investors including Greenoaks and Eclipse.

⚡️ Google Gemini ramps up their free tier

What happened: Google is making widespread updates to Gemini’s free tiers, adding support for multiple language and countries.

The details:

• The free tier of Gemini, called Gemini Flash is now available on web and mobile in 40 languages across 230 countries.

• The big benefit to Flash is the cost savings… it’s cheaper and more efficient. One of the biggest costs to running an AI is the power usage.

• Google is also upgrading Gemini to Gemini’s context window to 32,000 tokens (24,000 words). This is the input data the AI considers before generating the desired output. Such as a prompt or reference document.

• File uploads will soon be available for all Gemini users, regardless of tier. (Right now they’re available only for Premium members.)

• Gemini in messages rolled out in the European Economic Area, U.K., and Switzerland with new language support including French, Polish, and Spanish.

Why it matters: Free AI is the future of common use for average people. Especially AI integrated into Messages and other apps. Most people just need quick, simple AI uses. Flash might become the best in this category.

🔥 Mistral AI reveals newest model

What happened: Mistral revealed its new flagship AI model, Large 2, which competes 🥊 with top models from OpenAI and Meta.

The details: 

• Mistral claims Large 2 excels in code generation, mathematics, and reasoning, rivaling the latest models from OpenAI and Meta.

• According to Mistral (biased), Large 2 outperforms Meta’s Llama 3.1 405B in multiple areas despite having only 123 billion parameters. But only real use can determine that for sure.

• The model has a 128,000 token window, allowing it to process extensive data in a single prompt, equivalent to roughly a 300-page book.

• The model has been especially trained to minimize hallucinations, providing accurate responses and acknowledging its limitations.

• History: Mistral recently raised $640 million in a Series B funding round, led by General Catalyst, at a $6 billion valuation.

• Unlike Llama 3.1, Large 2 is not open-source; commercial applications require a paid license.

• It cannot process text and images at the same time.

• The model can be tested for free on Mistral’s ChatGPT competitor, Le Chat.

Why this matters: This launch comes just a day after Meta’s Llama 3.1 405B release. Llama 3.1 is literally a game changer, making the AI frontier open-source. But Mistral’s Large 2 is extremely promising. The more options and competition, the better.

😂 Memes galore!

Source: Midjourney

Give us your feedback!

Got feedback? Rate today’s newsletter by clicking below! 

Got more personal feedback? Fill in the box when you rate us and give us your criticism and feedback.

Thank you for reading! 

❤️Share the (ARTIFICIAL) Love!

Got a friend who wants to appear as smart as you? An AI fanatic or someone shaking in their boots of getting replaced? Share us by clicking the link below and you’ll get our love ;)