🤯 OpenAI’s GPT-4.5 Passes Turing Test

Also: Midjourney Launches V7 Image Model

Welcome, AI enthusiasts

Can you tell the difference between a human and a chatbot? Turns out, it’s getting harder. OpenAI’s latest model, GPT-4.5, fooled people 73% of the time when it played a specific persona, passing a version of the famous Turing test. Let’s dive in!

In today’s insights:

  • OpenAI’s GPT-4.5 Passes Turing Test

  • Midjourney Launches V7 Image Model

  • Amazon's AI Shops Anywhere for You

Read time: 4 minutes

LATEST DEVELOPMENTS

Source: Getty/Futurism

Evolving AI: OpenAI's GPT-4.5 convinced evaluators it was human 73% of the time when adopting a persona.

Key Points:

  • OpenAI's GPT-4.5 convinced evaluators it was human 73% of the time when adopting a persona.

  • GPT-4o and older models like ELIZA performed significantly worse.

  • The use of personas drastically improved AI's perceived humanness.

Details:

In 1950, Alan Turing proposed a test to measure machine intelligence. A human judge chats with both a person and a machine — without knowing which is which. If the judge can’t reliably tell them apart, the machine is said to have passed the test — meaning it’s good at pretending to be human. OpenAI's GPT-4.5 recently passed the Turing test, appearing human to evaluators at a remarkable rate of 73% when adopting a specific persona. Without persona instructions, its success dropped dramatically to 36%. Meanwhile, GPT-4o managed only 21%, even trailing the decades-old ELIZA chatbot at 23%. These findings reveal that AI convincingly mimics human interaction, particularly when given detailed, human-like roles to embody.

Why It Matters:

Passing the Turing test at 73% shows that AI can convincingly mimic human interaction — but only with clear instructions. This reveals AI’s power and limits, raising important questions: How human-like do we really want AI to become?

Evolving AI: Building an AI business? Becoming ISO 42001 compliant shows your customers that you are taking the necessary steps to ensure responsible usage and development of AI.

With Vanta, achieving compliance can be done in a fraction of the time. 95% of required document templates are pre-built, accelerating the work and helping you build trust.

Start with this free checklist — it breaks down the compliance process step by step.

Download the checklist now →

Evolving AI: Midjourney just rolled out V7, its first new AI image model in nearly a year.

Key Points:

  • V7 features personalized image generation, tuned by rating 200 images.

  • Improved image quality, smarter text prompts, and better coherence in details.

  • Includes new Draft Mode for faster, cheaper image renders.

Details:

Midjourney's V7, now in alpha, introduces a major upgrade with built-in personalization. Users rate around 200 images to train the model to their visual preferences. The model has better text prompt understanding, enhanced textures, and more accurate representations of bodies and objects. V7 offers two modes: Turbo and Relax, plus a Draft Mode that quickly creates lower-quality images, which can then be refined at a reduced cost. However, some features like image upscaling are still pending, with an update expected soon.

Why It Matters:

Midjourney’s V7 release shows the company is still competitive after OpenAI's recent viral success with ChatGPT's image generator. This could prompt other players, especially Midjourney's biggest rivals, to roll out new image generation models soon.

Source: Amazon

Evolving AI: Amazon is testing a new AI feature called "Buy for Me," letting users buy items from other sites without leaving Amazon.

Key Points:

  • Amazon’s "Buy for Me" searches and buys products from third-party websites directly within the Amazon app.

  • Powered by Amazon Nova and Anthropic’s Claude AI models, the agent fills out payment and shipping details autonomously.

  • Unlike other AI shopping tools, Amazon uses encryption so it can’t see external purchase details.

Details:

Amazon's latest AI feature ‘Buy for Me’ automatically finds and purchases items from external sites when Amazon itself doesn't stock the product. Using advanced AI models including Amazon Nova and Anthropic’s Claude, the agent completes transactions by autonomously providing your payment and shipping details. Amazon promises secure encryption, ensuring it doesn't track external purchases—a notable privacy approach compared to competitors like OpenAI and Google. However, the convenience comes with risks, such as potential AI errors in ordering and less user control over purchases.

Why It Matters:

Amazon’s new AI could change online shopping by simplifying cross-site purchases, but will shoppers trust AI enough to let it handle their money and orders without direct oversight?

 👀 Click on the image you think is real

QUICK HITS

📉 Microsoft reportedly pulls back on global data center expansion.

📃 U.S. Copyright Office issues highly anticipated report on copyrightability of AI-generated works.

🚀 ChatGPT's image generation explodes with 700M creations in first week.

🏭 Africa’s first ‘AI factory’ could be a breakthrough for the continent.

👀 Google's NotebookLM AI Can Now 'Discover Sources' for You.

📽️ Adobe launches Premiere Pro’s generative AI video extender.

📈 Trending AI Tools

  • 🗣️ Intellisay - Create a daily plan with your voice (link)

  • 🚀 Outreach - AI-powered selling (link)

  • 📄 aiPDF - An AI-powered tool for summarizing documents (link)

  • 🌐 NetMind Power - A decentralized ML and AI platform (link)

  • 🔍 Legal Graph - A tool to visualize connections between legal concepts, cases, statutes, and regulations (link)

  • 💼 AIJobs - A comprehensive platform that revolutionizes the job search and recruitment process in the AI industry (link)

Reply

or to participate.