TOGETHER WITH SURFSHARK
Welcome to AI Tool Report!
Friday’s top story: After 7 months of waiting, OpenAI has finally launched real-time vision capabilities for ChatGPT.
🏄♂️ To stay safe online, we recommend you try Surfshark: It hides your identity and secures all your devices from hackers, allowing you to surf the web securely.
🌤️ This Morning on AI Tool Report
👀 OpenAI launches real-time vision for ChatGPT
🏄♂️ How to become invisible to hackers with Surfshark
💦 Meta fights 4x deepfake rise
💼 How to build a business as an AI consultant with Innovating with AI
🧪 How to test personalization algorithms using ChatGPT
➗ Microsoft launches Phi-4: Best at Math?
📚 Harvard releases biggest-ever AI dataset
Read Time: 5 minutes
FACT OF THE DAY
🤔 According to McKinsey, by 2030, AI agents will replace 70% of office work and Goldman Sachs predicts they will add $7T to the global economy.
STOCK MARKETS
🌋 AI and tech stocks consolidated during a volatile trading session on Thursday. ETF’s such as the Nasdaq and S&P 500 are trading at all-time highs, so profit-taking is to be expected. Learn more.
— — — — — — —
VISION
Our Story: For day six of OpenAI’s “12 Days of OpenAI,” OpenAI has released Advanced Voice Mode with Vision—seven months after it first teased the feature—which means ChatGPT can now ‘’see’’ what users can, and respond in real-time.
🔑 Key Points:
Plus, Team, and Pro ChatGPT users, can now point their mobile camera at an object and have real-time conversations with ChatGPT about it, and they can also ask it questions about what’s on their device screen, too.
In a demo, the feature correctly identified and gave feedback on an anatomical drawing—“The location is spot on…the brain is more oval”—but it did get a geometry question wrong, suggesting it’s prone to hallucinations.
Not all Plus, Team, and Pro users will get access though, OpenAI couldn’t establish when EU, Switzerland, Iceland, Norway, or Liechtenstein users would get it, and Enterprise and Edu users will not get it until January.
🤔 Why you should care: Although it was first announced back in May—and has been delayed several times—OpenAI’s launch of real-time vision analysis capabilities comes as Google launched Gemini 2.0 yesterday—which can process visual and audio and has more agentic capabilities—which is powering a real-time video analysis feature, Project Astra (released to a small group of testers), its web-based task agent, Project Mariner, and its coding agent, Jules.
➕ As a little stocking filler, OpenAI has also added a preset Santa voice to its Advanced Voice Mode feature in ChatGPT (see ‘Recommended Resources for a demo), although no one knows if they got permission from Santa to use his voice…
— — — — — — —
TOGETHER WITH SURFSHARK
Our Report: Surfshark is an industry-leading VPN provider and security toolkit that hides your identity from hackers on the web, preventing them from stealing your info.
🔑 Key Points:
With just one click, Surfshark will give you full security on all your devices, in over 100 countries, meaning you can surf the internet safely and securely on any device, wherever you are.
With best-in-class security and privacy features—including data breach alerts, encryption, and malware protection—your information will remain secure, at all times.
Plus, as an AI Tool Report reader, you get 4 months more when you get a 2-year subscription, so you can remain invisible to hackers, all year round, without any worries.
🤔 Why you should care: For just $1.99 p/m (87% less than other VPN providers) Surfshark will hide your identity and secure all your devices from hackers, allowing you to surf the web, in any location, securely.
TRENDING TOOLS
Meco is a distraction-free space for reading and discovering newsletters, separate from the inbox ⭐⭐⭐⭐⭐ / 5 (Product Hunt)
SciSummary summarizes scientific articles using AI
Zipy debugs native mobile apps with AI
— — — — — — —
DEEPFAKES
Our Report: To help combat the 4x increase in AI deepfakes over the last year, Meta has released a new watermarking tool—Video Seal—that applies imperceptible watermarks to AI-generated videos to detect when AI has been used, to “protect originality," which compliments its previously launched watermarking tools: ‘Watermark Anything’ and ‘Audio Seal’
🔑 Key Points:
Video Seal is open source, meaning it can be easily integrated into existing software, and alongside the watermarking, it can also add invisible messages to videos, to identify their origins.
The watermarks are also resilient to common editing techniques, like blurring, cropping, and compression, but Meta did say that excessive editing or compression could damage the watermark, making it unidentifiable.
Although there are other video watermarking tools available—Google’s SynthID, for example—Meta believes that “many methods originate from image watermarking, which is not suitable for video.”
🤔 Why you should care: The biggest challenge for Meta, though, is persuading unmotivated developers and industry professionals to use their new tool (especially if they’re already using other solutions), so, to help with this, they’re launching a benchmarking leaderboard—Meta Omni Seal Bench—that compares the performance of different watermarking tools, and running a seminar about watermarking technology at next years International Conference on Learning Representations (a major AI-focused conference).
— — — — — — —
TOGETHER WITH INNOVATING WITH AI
The AI consulting market is about to grow by a factor of 8X – from $6.9B to $54.7B in 2032.
But how does an AI enthusiast become an AI consultant?
How well you answer that question makes the difference between just “having AI ideas” and being handsomely compensated for your contribution to an organization’s AI transformation.
Thankfully, you don’t have to go it alone – our friends at Innovating with AI have welcomed 300 new students into The AI Consultancy Project, their new program that trains you to build a business as an AI consultant.
Some of the highlights current students are excited about:
The tools and frameworks to find clients and deliver top-notch services
A 6-month plan to build a 6-figure AI consulting business
Students getting their first AI client in as little as 3 days
And as a reader of AI Tool Report, you get early access to the next enrollment cycle.
PROMPT ENGINEERING
— — — — — — —
Friday’s Prompt: How to test personalization algorithms using ChatGPT
Type this prompt into ChatGPT:
Generate a framework for A/B testing to evaluate the effectiveness of different personalization algorithms
Results: After typing this prompt, you will get an A/B testing framework that will allow you to test the effectiveness of different personalization algorithms.
P.S. Use the Prompt Engineer GPT by AI Tool report to 10x your prompts.
ACTIONABLE INSIGHTS
❗Did you miss the AI Skill Sprint, on Skool, last month? Attendees mastered 6 crucial AI skills in just 6 weeks.
Here’s a taster of what you missed.
✅ What We Covered: Course 1 was about finding ways to assess and enhance an organization’s readiness for AI integration. Ben Etherington, from Founders Forge, gave practical ways to align AI with business strategy and explained how to implement them across all areas of business, effectively.
Want to see what else we covered during the 6-week skill sprint? Subscribe to The AI Report and get access to all 6 sprints.
Want to join the next one? It’s coming in 2025, so be ready to sign up here.
BREAKING NEWS
— — — — — — —
LAUNCHES
Microsoft has launched Phi-4—the smallest AI model in its Phi family—that reportedly outperforms its predecessors and rivals (including Gemini Pro 1.5) in several areas, especially mathematical reasoning.
This is because Microsoft has used high-quality synthetic datasets to train Phi-4, along with “high-quality datasets of human-generated content” and unspecified “post-training innovations.”
Using synthetic data and improved post-training methods is something many AI companies are looking into, to overcome the reduced availability of unused, organic data sources, such as web content or code.
— — — — — — —
DATA
Harvard University is releasing a dataset—for training AI models—that contains over 1M public domain books, and is 5x the size of the notorious Books3 dataset (used to train high-profile models like Meta’s Llama).
OpenAI and Google are funding the project—which is run by Harvard’s Institutional Data Initiative—and the dataset will contain books, scanned by Google Books, that are no longer protected by copyright.
The dataset will also span different genres, decades, and languages, with content ranging from Shakespeare, Dickens, and Dante classics, to obscure Czech textbooks and Welsh dictionaries.
RECOMMENDED RESOURCES
Here’s a little teaser 👀
As some of you guessed in last week’s poll, we’re excited to announce that AI Tool Report is rebranding! 🎉
We’re evolving into
‘The AI Report by Upscaile.’
Upscaile is our parent company—delivering custom AI education at an Enterprise level—and our rebrand marks a shift towards providing you with a more professional outlook and AI skills, training, and education that you can practically apply to your day-to-day lives.
To help us deliver this, in the New Year, we’ll be introducing Amanda Greenwood as our new Editor-in-Chief of The AI Report. Any feedback, send it her way!
In the meantime, stay tuned for the big re-brand reveal….
We read your emails, comments, and poll replies daily.
Hit reply and tell us what you want more of!
Got a friend who needs to learn more about AI? Sign them up to the AI Tool Report, here.
Until next time, Martin & Liam.
P.S. Don’t forget, you can unsubscribe if you don’t want us to land in your inbox anymore.