• Synthetic
  • Posts
  • Will We Finally Talk to Dolphins with AI? 🐬

Will We Finally Talk to Dolphins with AI? 🐬

Plus, Meet Your New Robot Butler in Training 🤖🍸🍽️

Author

Steve Brown
April 17, 2025

Subscribe to Synthetic

The week’s most interesting and relevant AI news and analysis

This Week in AI

The Human-centered Artificial Intelligence department at Stanford University released its annual detailed report on the state of AI and the AI industry. The 455-page report is filled with charts, diagrams, and data on AI trends and insights. Top takeaways include:

  • AI performance on demanding benchmarks continues to improve.

  • AI is increasingly embedded in everyday life.

  • Research shows substantial productivity impacts, so businesses are all in on AI, fueling record investment and usage.

  • The U.S. still leads in producing top AI models, but China is closing the gap.

  • Global AI optimism is rising, but deep regional divides remain.

  • AI is becoming more efficient, affordable, and accessible.

  • Governments are stepping up on AI with regulation and investment.

  • AI and computer science education are expanding, but gaps in access and readiness persist.

  • Industry is racing ahead on AI, and the frontier is tightening.

  • AI is having a substantial impact on scientific research.

  • Complex reasoning remains a challenge.

OpenAI has been on a tear lately. Last week (just in time to miss coverage in Synthetic), it released its latest non-reasoning model, GPT-4.1. This multimodal model comes in three sizes: GPT-4.1, GPT-4.1 Mini, and GPT-4.1 Nano. The smaller models are cheaper to use and deliver rapid responses.

The new reasoning models, o3 and o4-mini, have excited the AI community. As described in a previous issue of Synthetic, reasoning models can break down complex challenges into more manageable tasks and reason their way through problems, making them more useful and improving accuracy. While they don’t remove hallucinations altogether, they are more reliable than previous models. The other big deal with o3 and o4-mini is that they can access and use tools to get things done, making them agentic. OpenAI describes o3 as their “most powerful reasoning model that pushes the frontier across coding, math, science, visual perception, and more,” and o4-mini as “a smaller model optimized for fast, cost-efficient reasoning.”

Quick Hits

Video: Meet NEO, Your Robot Butler in Training 🤖

The CEO of Norwegian robotics company 1X delivered this TED Talk last week at TED 2025. Bernt Børnich explains why 1X decided to go after the home market first before the industrial segment that most of his competitors have prioritized. (15m)

AI Tech and Innovation

Source: Google

Google has developed a new AI model, DolphinGemma, to decipher how dolphins communicate with intricate clicks, whistles, buzzes, and squawks. The model is small enough to run on a smartphone, making it easy for researchers from the Wild Dolphin Project to gather data in the field. 🐬

Microsoft has surprised the industry by releasing a new ultra-compact language model that performs competitively in math, coding, and conversation. Typically, large language models use 8-bit, 16-bit, or 32-bit precision to render tokens, but this new model uses only 1.58 bits per weight, allowing it to run efficiently on GPUs and CPUs. The model, imaginatively named BitNet b1.58 2B4T, has been released as open source on Hugging Face. It was trained on a dataset of 4 trillion tokens and is built on a new architecture adapted from the traditional transformer architecture that underpins most other large language models.

The newest trend in programming is “vibe coding,” which involves creating code using human language prompts. Leading AI researcher Andrej Karpathy coined the term in a post where he suggested coders “fully give in to the vibes, embrace exponentials, and forget that the code even exists.” Natural language coding is the next natural step for software programmers. Early coders programmed computers in binary or machine code. This was unwieldy and difficult work so compilers and other tools were built so programmers could use simple languages like Fortran, BASIC, and COBOL to create code. Over time, higher-level languages like C++ and Python emerged to make it easier for coders to build complex software and work together in teams. Today, AI allows almost anyone to write software and become a coder, explaining what they want in natural human language. 👩🏽‍💻

AI Insights

According to McKinsey and as graphed by the Visual Capitalist, U.S. data center energy use is forecast to quadruple between 2023 and 2030.

In a separate analysis, AI investor and friend of Synthetic Pankaj Kedja suggests the energy industry may need to invest $4 trillion to meet expected demand for AI. Today, AI data centers worldwide consume 10GW of power, half of which are in the U.S. The industry expects we will need 125GW to power AI data centers by 2030, with two-thirds of those in the U.S., representing a 12x increase! Assuming an average investment of $35B per GW of generating capacity, we must spend $4 trillion to meet demand. ⚡️

Alibaba co-founder Jack Ma has emphasized the importance of developing AI technology that focuses on collaboration and enhances human capabilities rather than overpowering them. The tech visionary advocates a symbiotic relationship that leverages human creativity and machine efficiency to tackle global challenges. Ma has called for regulating AI development to prevent misuse and make AI use more transparent.

Tools To Try

This AI-powered software helps you find relevant journalists and podcasters and quickly build the perfect pitch. Get help with press mentions, build authority backlinks, establish thought leadership, and improve your Google page rankings with brand mentions. Connect with over 700K journalists and podcasters. Free trial!

A copilot for your sales force that surfaces fresh leads, maximizes buyer engagement, and boosts email open rates by up to 50%. Empower your sales reps, uncover new opportunities, and grow revenue with an all-in-one, multi-channel AI platform that elevates the capabilities of your sales team—free 14-day trials.

Make a logo in minutes—no design skills required! Auto-create a brand kit with fonts and colors to complement your new logo and create a brand identity. Instantly create branded marketing materials, then use over 300 templates to customize branded business cards, social profiles, email signatures, and more.

Maintain healthy calendar habits, improve your productivity, optimize cross-team meetings, boost collaboration, and improve work-life balance. Reclaim schedules 1:1s and defends your calendar so you can focus. Syncs multiple calendars and integrates with Google Calendar, Zoom, Slack, and HubSpot.

Incorporate your company, access one-click growth tools, stay compliant, and manage everything your business needs — all online, from anywhere. Launch your U.S. business in minutes with no paperwork or legal headaches.

Need to expand internationally, but not sure where to start? Oyster makes it easy to find, hire, and retain local talent. They handle the details (country-specific labor laws, international tax laws, compliance, and global payroll) so you can focus on finding the right talent and ramping their impact.