Greetings AI enthusiasts. Google, in collaboration with Georgia Tech and the Wild Dolphin Project, is developing a large language model called DolphinGemma that could bring researchers closer to two-way communication with dolphins.

In today’s email:

• Google's AI Learns to Communicate with Dolphins

• OpenAI Introduces GPT-4.1

• As AI Grows, Gen Z Looks to Schools and Workplaces for Direction

LATEST NEWS

GOOGLE

🐬 Google's AI Learns to Communicate with Dolphins

AI Spotlight: Google, in collaboration with Georgia Tech and the Wild Dolphin Project, is developing a large language model called DolphinGemma that could bring researchers closer to two-way communication with dolphins. By analyzing and generating dolphin-like vocalizations, this AI-based approach aims to uncover patterns in dolphin communication and potentially test how dolphins respond to new, artificial "words."

Key details:

• Researchers used decades of dolphin audio data to train DolphinGemma, enabling it to produce realistic dolphin sounds, including complex burst pulses.

• The team plans to test how dolphins react to AI-generated vocalizations using wearable technology that plays these sounds while humans interact with the animals.

• Experts caution that dolphins might imitate sounds without truly understanding them, and that current data does not confirm whether dolphins use symbolic language.

• Variability in dolphin whistles and the lack of clear wordlike structure could pose challenges for AI models that rely on consistent symbolic patterns.

While true dolphin-human conversation may remain out of reach, the project highlights AI’s potential to shed new light on dolphin cognition. Even if language remains unproven, deeper insights into dolphin behavior could reshape how we view and protect these intelligent animals.

OPENAI

🖥️ OpenAI Introduces GPT-4.1

AI Spotlight: OpenAI has introduced a new family of models: GPT‑4.1, GPT‑4.1 mini, and GPT‑4.1 nano that build on the strengths of previous versions with sharp improvements in coding, instruction following, and long-context comprehension. These models are designed to meet real-world developer needs while delivering faster performance and lower costs.

Key details:

• GPT‑4.1 shows leading performance in coding, scoring 54.6% on SWE-bench Verified, a notable improvement over previous models.

• It outperforms earlier versions in instruction following, scoring 38.3% on Scale’s MultiChallenge benchmark.

• The models support a massive 1 million token context window and achieve state-of-the-art results in long-context benchmarks like Video-MME.

• GPT‑4.1 mini and nano offer competitive performance at reduced latency and cost, with nano excelling in fast, lightweight tasks like classification.

The GPT‑4.1 model family raises the bar for AI model efficiency and utility, especially for developers creating agents or tools that handle complex tasks with minimal supervision. As OpenAI phases out GPT‑4.5 Preview, these new models are set to become the go-to tools for scalable, intelligent applications.

GEN Z

🌎️ As AI Grows, Gen Z Looks to Schools and Workplaces for Direction

AI Spotlight: A new Gallup survey reveals that while Gen Z frequently engages with generative AI, many feel conflicted about its impact and want clearer guidance on how to use it effectively. The results highlight the need for structured support in both educational and professional settings.

Key details:

• Widespread Use with Mixed Emotions: Nearly 80 percent of Gen Z uses AI, but over half of adults in this group report feeling anxious about it.

• Desire for Education and Guidance: Young people want clear rules on using AI in school and more instruction on how to apply it in real-world settings.

• Concerns About Cognitive Impact: Respondents are twice as likely to say AI harms rather than helps their critical thinking.

• Skepticism About AI in Professional Settings: Despite using AI at work, many Gen Z adults distrust AI-generated work and prefer human input.

Gen Z’s frequent use of AI is paired with a strong desire for better guidance, reflecting a generation eager to harness AI responsibly while guarding against its potential downsides. Both schools and employers have an opportunity to meet this demand by building thoughtful, transparent frameworks for AI education and use.

MORE STORIES IN AI

AI-boosted cameras help blind people to navigate
Adobe takes stake in Synthesia, startup behind AI clones for corporate videos
Meta to use public posts, AI interactions to train models in EU
New AI Model Predicts 44 Earth-Like Planets — And It’s 99% Accurate
Apple details how it plans to improve its AI models by privately analyzing user data