- AI PlanetX
- Posts
- Amazon Voice AI Beats GPT-4o
Amazon Voice AI Beats GPT-4o
Google Unveils First-Ever AI TPU

Welcome to another edition of AI PlanetX.
Amazon debuts Nova Sonic to rival OpenAI voice; Google reveals Ironwood AI chip; Murati recruits OpenAI veterans.
Inside This Edition: đź’Ž
Hottest AI News
Top AI & SaaS Tools
Top AI & Tech News
Interesting Uses of AI: Create Anime from a Single Image
AI Video Tutorial
F-R-E-E AI Course of the Day: Intro to Language Models in Python
Hottest AI News
Amazon
Amazon Introduces Nova Sonic to Rival OpenAI Voice Models

Amazon has launched Nova Sonic, a generative AI voice model that generates natural-sounding responses. It is said to rival models from OpenAI and Google, marking a major leap in conversational AI.
Details:
Nova Sonic boasts a 4.2% word error rate and responds in 1.09 seconds, outperforming GPT-4o. It’s 46.7% more accurate in noisy environments during multi-person interactions
Available via Amazon's Bedrock platform with a bi-directional streaming API, Nova Sonic is 80% cheaper than GPT-4o and powers the upgraded Alexa+ assistant
Unlike legacy voice assistants, Nova Sonic intelligently routes requests, handles interruptions, and adapts to natural conversation dynamics
This release is a key step in Amazon's AGI strategy, led by SVP Rohit Prasad, with plans to expand AI models across image, video, voice, and other sensory inputs.
Run IRL ads as easily as PPC
AdQuick unlocks the benefits of Out Of Home (OOH) advertising in a way no one else has. Approaching the problem with eyes to performance, created for marketers with the engineering excellence you’ve come to expect for the internet.
Marketers agree OOH is one of the best ways for building brand awareness, reaching new customers, and reinforcing your brand message. It’s just been difficult to scale. But with AdQuick, you can plan, deploy and measure campaigns as easily as digital ads, making them a no-brainer to add to your team’s toolbox.
You can learn more at AdQuick.com
DeepMind
Google Unveils Ironwood as Most Powerful AI Accelerator Yet

Google unveiled Ironwood, its seventh-generation TPU, at the Cloud Next conference this week. Ironwood is the first TPU optimized for AI inference, focusing on running models rather than training them.
Details:
Ironwood delivers 4,614 TFLOPs of computing power, 192GB of RAM, and nearly 7.4 Tbps bandwidth. It will be available in 256-chip and 9,216-chip configurations later this year
The chip features an enhanced SparseCore for advanced ranking and recommendation tasks, reducing data movement and latency, making it Google’s most powerful and energy-efficient TPU
Ironwood enters a competitive AI accelerator market, with Nvidia leading, while Amazon and Microsoft are developing their own solutions
Google will integrate Ironwood with its AI Hypercomputer in Google Cloud. VP Amin Vahdat called it a breakthrough in inference, combining more power, memory, networking, and reliability.
Thinking Machines Lab
Mira Murati Attracts Key OpenAI Veterans to AI Startup

Mira Murati’s venture, Thinking Machines Lab, has added former OpenAI researchers Bob McGrew and Alec Radford to its advisory team as it develops customizable AI systems.
Details:
Startup’s website updated in March to show McGrew and Radford as advisers. McGrew was OpenAI's chief research officer until September 2024, while Radford worked on GPT, ChatGPT, DALL-E, and Whisper before leaving in late 2023
Thinking Machines Lab aims to create more customizable, understandable, and capable AI systems, focusing on user-tailored tools, though product details are unclear
Leadership includes Murati as CEO, John Schulman as chief scientist, and Barret Zoph as CTO. The company has hired dozens from leading AI labs, including OpenAI and Google DeepMind
Murati, who left OpenAI in October after leading ChatGPT, DALL-E, and Codex, is seeking over $100 million in funding for Thinking Machines Lab.
Top AI & SaaS Tools
SellerPic (Life-time Deal): Use AI to transform product photos into dynamic visuals, model shots, and lifestyle scenes, including videos
Deep Cogito: Family of lightweight open-source AI models with adjustable reasoning capabilities, surpassing Meta’s Llama in performance [F-R-E-E]
Clarisign: Manages the entire contract workflow—from drafting to signing—by flagging risks, auto-filling key details, reducing paperwork
Genspark: Autonomous agent built to tackle real-world tasks—like making restaurant calls with a lifelike synthetic voice [F-R-E-E]
EZsite: Build websites/apps with auto-saving to your database, capture leads with built-in chatbots, and boost sales using AI automations [F-R-E-E]
Top AI & Tech News
Stanford’s 2025 AI Index shows AI got smarter and cheaper—but rising costs, safety gaps, and fierce competition loom
Non-profit Sage Future used its AI agent village in an experimental test to raise $257 for charity
Snap launched an AI ad format using “One Snap” tech, letting users insert themselves into brand content via phone’s front camera
NVIDIA released Nemotron-Ultra, a 253B open-source model that outperforms DeepSeek R1 and Llama 4 Behemoth on key benchmarks
Artisan raised $25 million in Series A funding while still hiring humans despite its "Stop Hiring Humans" campaign
Interesting Uses of AI
AI Art Spotlight

Model: Midjourney V7
Prompt:
a cinematic poetic portrait of a girl in the nature, eyes opened, in a lot of different flowers, nature and bokeh ine the background --ar 3:4 --raw --stylize 500 --v 7
AI Prompt Tip of the Day
Create anime from a single image.
With Gen-4 and ChatGPT, you can turn any image into a multi-scene cartoon.
Here’s how 👇
— Framer 🇱🇹 (@0xFramer)
3:23 PM • Apr 7, 2025
Top AI Video Tutorial
ChatGPT-4o replaces your $10,000 Ad Agency (Full Tutorial)
Complimentary AI Course of the Day
Intro to Language Models in Python

Learn to build and use common language models in data science, including bag-of-words, tf-idf, and word embeddings. Gain essential skills for working with advanced AI language models.