AI Inside: New Week, New Seek
For Wednesday, March 12, 2025
Well hello! First, a huge thank you to our patrons for the increasing support! We are building something with AI Inside!
OH and Intel agrees. Check this out from my wanderings on the show floor at Mobile World Congress. I was taken aback when I looked up to find this staring at me! We are validated!
Big thanks to wireman92002, this week’s Patron of the week. And OF COURSE our amazing roster of Executive Producers:
DrDew, Jeffrey Marraccini, WPVM 103.7 in Asheville NC, Dante St James, Bono De Rick, Jason Neiffer, and Jason Brady. WOW, we are so grateful for you all!
Alright onto this week’s episode of the podcast!
Jason’s conversation with Sameer Samat, Head of Android at Google
I sat down with Sameer Samat in a Google conference room tucked away deep in the halls of Mobile World Congress last week in Barcelona, Spain. I mention dripping sweat at the top of the interview primarily because not 5 minutes before I hit record, I was still very lost amidst the noise and chaos of the show floor. Sameer was incredibly gracious when I arrived a few minutes late for our chat, we sat down, and dove right in!
In this interview, Sameer and I talk all about the new Gemini Live features soon to hit Android smartphones that bridge the gap between what our smartphones can do now in the way of multimodal conversational interaction, and what next generation systems like Project Astra will be able to do with the smartglasses form factor once they launch.
We also discuss the fears of developers in an age of increased agentic AI on smartphones. Primarily, I ask whether developers should be concerned with agents making apps less of a priority for Google and mobile phone users.
But an overall theme of the conversation is something that Jeff and I have discussed many times on the show: When do we reach a point when the technology itself (AI) takes a back seat to the benefits of its existence? So much of the messaging coming from companies like Google is focused on the technology, the fact that AI is driving so much change and innovation. But users really want to know more about how their devices can really improve their lives. Knowing that AI is driving the experience matters a lot less and in some ways can give them a sour taste in their mouths before ever trying the new product or service.
The full interview can be viewed in today’s episode. I also wrote an article on these topics for ZDNet that has published today, so check that out as well!
OpenAI's New AI Agents Tools
OpenAI has introduced new tools for developers and businesses to build their own AI agents, part of the Responses API set to replace the Assistants API within the next year. These tools include capabilities for AI agents to control computers, similar to OpenAI's Operator agent, though they currently have a 38.1% success rate on complex tasks. Sam Altman believes 2025 will be the year agents "join the workforce," but for now, they continue to promise more than they can deliver. OpenAI acknowledges that these tools are still in the early stages and have much room for improvement.
Manus: The New Chinese AI Model
Manus, a new Chinese AI model available on Hugging Face, is gaining significant attention. Developed using existing models like Claude and Alibaba's Qwen, Manus can perform multistep tasks, though examples so far haven't shown consistent completion of these tasks. The launch had a similar energy to it as the recently released DeepSeek model, with invite codes selling for thousands in Chinese reseller apps. The research lead describes Manus as a "completely autonomous agent" that bridges conception and execution, promising a new paradigm in human-machine collaboration. (this all sounds very familiar!)
Apple's AI-Driven Siri Delays
Apple is delaying the release of a series of AI-driven features for Siri by a year, impacting plans for a new home device that combines a HomePod and iPad. This device, envisioned as a "wall tablet," was expected to be priced between $130 and $230. The delay affects Apple's App Intents technology, which allows AI to control other applications, similar to Google's Gemini extensions.
Larry Page's Dynatomics
Google co-founder Larry Page is secretly building Dynatomics, a company focused on applying AI to product manufacturing. Dynatomics aims to create highly optimized designs for factories to build. Chris Anderson, formerly CTO of Page's Kittyhawk startup, is leading the effort. Details about the company are scarce, but it represents a significant venture into AI-driven manufacturing. Also, what can I say, I’m curious about anything Larry Page decides to focus his time on.
AI Search Engines Struggle with Citations
A study by the Columbia Journalism Review compared eight AI search engines and found them lacking in accurately citing news sources. Chatbots often provided incorrect or speculative answers, with premium versions being more confident in their inaccuracies. They also disregarded Robot Exclusion Protocol preferences and created fake links. This highlights a common issue with AI tools being "confidently wrong."
AI in Creative Writing
Sam Altman has expressed admiration for a new AI model capable of writing metafictional literary short stories about AI and grief. This model, still in development, has impressed Altman with its grasp of metafiction. There is no release date announced for this model, but it represents a promising advancement in AI's creative writing capabilities. Go ahead and read the creative writing that Sam seems so smitten by.