Google working to decode dolphin communication using AI
Dolphins are one of the smartest animals on Earth and have been revered for thousands of years for their intelligence, emotions and social interaction with humans.
Now Google is using artificial intelligence (AI) to try and understand how they communicate with one another – with the hope that one day humans could use the technology to chat with the friendly finned mammals.
Google has teamed up with researchers at Georgia Institute of Technology and the Wild Dolphin Project (WDP), a Florida-based non-profit which has been studying and recording dolphin sounds for 40 years, to build the new AI model called DolphinGemma.
'SUPER POD' OF MORE THAN 1,500 DOLPHINS OFF CALIFORNIA COAST CAPTURED ON DRONE VIDEO
For decades, WDP has correlated sound types with behavioral contexts. For instance, signature whistles have been used by mothers and calves to reunite, while burst pulse "squawks" are often observed during dolphin fights, researchers said, according to a Google blog on the project.
Click "buzzes" are often used during courtship or chasing sharks.
Now, using the vast data gathered by WDP, Google has built DolphinGemma, building upon Google's own AI lightweight open model, known as Gemma.
DolphinGemma has been trained to analyze the vast library of recordings to detect patterns, structures, and even potential "meanings" behind the dolphin communications or vocalizations.
Over time, DolphinGemma will try to organize the dolphin sounds into categories — almost like words, sentences, or expressions in human language.
CHINESE HUMANOID ROBOT WITH EAGLE-EYE VISION AND POWERFUL AI
"By identifying recurring sound patterns, clusters and reliable sequences, the model can help researchers uncover hidden structures and potential meanings within the dolphins' natural communication — a task previously requiring immense human effort," a post on Google about the project reads.
"Eventually, these patterns, augmented with synthetic sounds created by the researchers to refer to objects with which the dolphins like to play, may establish a shared vocabulary with the dolphins for interactive communication."
DolphinGemma uses Google's Pixel phone technology, specifically the audio recording technology used in Pixel devices, to make clean, high-quality sound recordings of dolphin vocalizations.
The Pixel phone technology can separate out dolphin clicks and whistles from background noise like waves, boat engines, or underwater static. That clean audio is critical for AI models like DolphinGemma, because messy, noisy data would confuse the AI, researchers said.
Google says it plans to release DolphinGemma as an open model this summer, allowing researchers around the world to use and adapt it.
Although it's trained on Atlantic spotted dolphins, the model could also help study other species like bottlenose or spinner dolphins with some fine-tuning, researchers said.
"By providing tools like DolphinGemma, we hope to give researchers worldwide the tools to mine their own acoustic datasets, accelerate the search for patterns and collectively deepen our understanding of these intelligent marine mammals," the blog post reads.
Hashtags
Try Our AI Features
Explore what Daily8 AI can do for you:
Comments
No comments yet...
Related Articles
Yahoo
an hour ago
- Yahoo
IBM, Google claim quantum computers are almost here after major breakthroughs: ‘It doesn't feel like a dream anymore'
The decades-long quest to create a practical quantum computer is accelerating as major tech companies say they are closing in on designs that could scale from small lab experiments to full working systems within just a few years. IBM laid out a detailed plan for a large-scale machine in June, filling in gaps from earlier concepts and declaring it was on track to build one by the end of the decade. 'It doesn't feel like a dream anymore,' Jay Gambetta, head of IBM's quantum initiative, told Financial Times. 'I really do feel like we've cracked the code and we'll be able to build this machine by the end of the decade.' Google, which cleared one of the toughest technical obstacles late last year, says it is also confident it can produce an industrial-scale system within that time frame, while Amazon Web Services cautions that it could still take 15 to 30 years before such machines are truly useful. Quantum computing is a new kind of computing that doesn't just think in 0s and 1s like today's computers. Instead, it uses qubits — tiny quantum bits — that can be 0, 1, or both at the same time. This lets quantum computers explore many possibilities at once and find answers to certain complex problems much faster than normal computers. Quantum computing could speed up the discovery of new drugs and treatments, make artificial intelligence systems faster and more capable and improve the accuracy of market predictions and fraud detection in finance. It could also dramatically improve efficiency in areas like traffic routing, shipping, energy grids and supply chains while driving green innovation by helping design better batteries, cleaner energy systems and more sustainable technologies. But scaling them up from fewer than 200 qubits — the quantum version of a computing bit — to over 1 million will require overcoming formidable engineering challenges. Qubits are inherently unstable, maintaining their special quantum states for only fractions of a second, and adding more of them can create interference that scrambles calculations. Even if the fundamental physics problems are solved, the industry still faces the task of industrializing quantum technology. This means building chips that can house large numbers of qubits, and developing much bigger refrigeration units to keep the systems at near absolute zero. Systems using superconducting qubits, like those from IBM and Google, have made some of the fastest progress but require extreme cooling and are difficult to control. Meanwhile, some companies are betting on radically new qubit designs. Amazon and Microsoft claim to have harnessed a new state of matter to produce more reliable components, although these are still in early development. 'Just because it's hard, doesn't mean it can't be done,' Mark Horvath, an analyst at Gartner, told FT.
Business Wire
3 hours ago
- Business Wire
Humanoid Robots: 'Vision and Reality' POSITION PAPER by IFR
FRANKFURT AM MAIN, Germany--(BUSINESS WIRE)-- Humanoids are considered to be the next big thing in robotics: China, the world's largest market for industrial robots, has set out specific targets for its plans to mass-produce humanoids. Meanwhile, tech companies in the US and Europe are announcing significant funding. The vision is to create general-purpose robots based on human motion mechanics. What are the trends, opportunities, and potential limitations of humanoids? The International Federation of Robotics has released a new positioning paper that provides valuable insights. Humanoid Robots: 'Vision and Reality' POSITION PAPER Released by Word Robotics Federation IFR Share 'Futuristic humanoids working in homes, businesses, and public spaces fuel people's interest," says Takayuki Ito, President of the International Federation of Robotics. 'Since our environment is optimized for the human body, the idea of a quick, universal helper to maintain manufacturing and services is evident. If and when a mass adoption of humanoids will take place remains uncertain. In any case, humanoids are not expected to replace current robot types in future. Instead, they will complement and expand upon existing technology.' Humanoids by region In the United States, tech companies like NVIDIA, Amazon and Tesla are heavily developing advanced AI and robotics technologies. Besides military funding, a lot of private investments support this development and result in a large number of start-ups developing humanoid robots. China put humanoids in the center of their national strategy. The government wants to showcase its competences and global competitiveness in this field of technology. One key element of the Chinese strategy is to establish a supply chain for key components that is scalable. Japan has been a pioneer in the development of humanoid robots, with Honda's Asimo being an early example unveiled in October 2000. The focus is on creating robots that can live harmoniously with humans and are accepted as part of society. Europe places a strong emphasis on the ethical implications of robotics and AI. Europe has a significant focus on collaborative robots that work alongside humans in industrial settings. Outlook Humanoids are well placed to automate complex tasks with which current robots struggle using traditional programming methods. However, mass adoption as universal household helpers may not happen in the near or medium term. About IFR
Forbes
5 hours ago
- Forbes
Inside Google's Plan To Use AI To Survey Americans On Their Political Views
E arlier this year, Scott Rasmussen, the longtime pollster and political commentator, traveled to Bowling Green, Kentucky, known for its Corvette assembly plant and for being the hometown of Fruit of the Loom. After he toured the city, about 65 miles north of Nashville, he returned home with a novel idea: use AI to transform polling, a notoriously fickle and imprecise discipline that he had studied for decades. To pursue the project, he teamed up with an unlikely partner — Google. Rasmussen had visited Bowling Green to check out the work of Jigsaw, a think tank inside Google that tackles big societal challenges like online disinformation. At the time, it was working with the local government in the Kentucky city and surrounding county on an experiment aimed at jumpstarting civic engagement. Jigsaw asked residents to answer questions about the issues they cared most about, from the potential arrival of a Dave & Buster's to the debate over marijuana legalization. From there, it would use a Google AI tool called Sensemaker, built from its language model Gemini, to analyze the answers and separate the residents' disagreements from where they had common ground. Rasmussen told Forbes was stunned by the results and said he saw in them an opportunity to try and repair our balkanized political discourse. Why not survey the entire country? The goal is to uncover common ground, said Rasmussen, who cofounded ESPN with his father, Bill, in 1979. He argues that the U.S. political population is not 50-50, but more 10-10-80: the 10% that is MAGA conservative is warring with the 10% on the far left, he said. He's hopeful that this project will highlight the rest. 'The 80% in between are trying to keep their heads down and avoid getting hit in the crossfire.' The problem with traditional polls, he said, is that closed-ended questions empower the asker to frame or tilt the discussion with yes or no answers, a binary that reduces nuance. 'When you begin to ask people the questions in a different way — or begin to address their opinions in a different way — you hear things you never thought to ask,' he said. The result is an ambitious project: As the United States turns 250-years-old next July, Jigsaw partnered with Rasmussen's Napolitan Institute, a nonprofit dedicated to the future of polling and analysis, on an initiative to use AI in a similar fashion to poll Americans about the future of the country. The project, called We The People, Google exclusively told Forbes , will gather five to 10 people each from all 435 congressional districts in the U.S. to answer questions about what it means to be an American, the most urgent issues facing the country, and where the nation might go from here. 'We want to use AI to give people voice and choice in the world around them,' Yasmin Green, CEO of Jigsaw and a 19-year veteran of Google, told Forbes . 'If people don't feel that they have a voice, or their voice matters to their policymaker, they don't feel that they are enfranchised or have agency.' 'The ability to get people to respond in their own language, to get them to respond to how other people have answered questions, it's revolutionary for the industry.' Scott Rasmussen The project has a sheen of feel-good unity to it, but its implications are further reaching. The opportunity for AI to transform political polling is profound. Today, polls can be flawed and inaccurate — they infamously predicted Hillary Clinton would win the 2016 election. They are typically categorized between campaign polling (whether or not you'll vote for a candidate) and public opinion polling (how you feel about a certain issue), but both types have a similar format. Pollsters call or text with surveys and hope people answer, then control for demographics. It might be more difficult for AI to help predict who someone might vote for, Rasmussen said, but it could still potentially transform how campaigns are organized. For example, someone running for political office could use AI-generated prompts to tease out topics of concern and importance for constituents, at a level far more in depth than the yes-or-no answers of traditional polls. Then the politician could craft a hyper-specific campaign based on those insights. 'The ability to get people to respond in their own language, to get them to respond to how other people have answered questions, it's revolutionary for the industry,' said Rasmussen. 'It's an entirely different game.' Green sees the broader potential too — listing market and public opinion research as applications. 'When people make a business out of your ideas, that's really validating. So I'd be happy to see that happen,' said Green. 'Our job is to validate that this is useful.' The project takes inspiration from a similar experiment in Taiwan. In 2016, the government decided to take policy debates to an online platform called vTaiwan, which let people discuss legislative recommendations — like how to regulate online sales of alcohol — and upvote them. 'Instead of a traditional poll or survey where people are in very fixed silos, these conversations are generative and deliberative in a sense that it allows people to come up with new ideas, new feelings that other people resonate with,' Audrey Tang, who helped implement the project before becoming Taiwan's first digital affairs minister, tells Forbes now. Since then, vTaiwan has helped facilitate around 20 legislative reviews, and in 2023, it received a $100,000 grant from OpenAI as part of a program to 'fund experiments in setting up a democratic process for deciding what rules AI systems should follow.' 'We want to use AI to give people voice and choice in the world around them.' Yasmin Green Compared to AI, the polling business is a relatively modest one. In 2025, the market, which includes Ipsos, Pew, Nielsen and Quinnipiac, grew to $8.93 billion, up slightly from $8.7 billion a year before, according to the firm Research and Markets. It's expected to steadily rise to $10.23 billion by 2029. Firms like Pew and Quinnipiac, run out of Quinnipiac University in Connecticut, specialize in public opinion polls, while the polling company that Rasmussen founded in 2003, Rasmussen Reports, focuses more on candidates. His ties to Rasmussen Reports, which he departed in 2013, could result in unwanted optics for the project due to controversies over the firm's alleged conservative bias. The company has long been accused of favoring Republican candidates, skewing toward older Americans and generating results that favor conservatives. During the 2024 U.S. election, the company reportedly shared polling results with officials of the Trump campaign. Asked about that alleged conservative bent, Rasmussen tries to distance himself from the firm he founded. 'I've had nothing to do with it since I walked out the door. They have nothing to do with this project, and it's really not an issue,' he said. The project comes as AI is already leaving fingerprints on politics and elections around the world. During the 2024 U.S. presidential race, a deepfake of President Joe Biden's voice told New Hampshire voters not to turn out for the state's primary. In Indonesia, the political party Golkar used AI to digitally recreate Suharto, a dictator who died in 2008, to endorse the party's candidates. President Donald Trump has repeatedly posted AI-generated memes and videos to his Truth Social platform. Last month, he shared one of former President Barack Obama being detained in the Oval Office. Jigsaw and the Napolitan Institute hope to deploy AI in a way that's more productive. The project will have three phases. During round one, participants will answer questions about freedom and equality. 'What does freedom mean to you?' for example. Then Google's AI model will generate follow up questions to dig deeper into the topic, adopting a 'Socratic persona,' Green said. In this case, it might ask, 'If feeling free means expressing yourself without judgment, could you share an example of a time when you felt most restricted?' The idea is to start broad and slowly get more specific. In round two, Google's AI tool will distill the responses from round one into overarching themes, discussion points and data visualizations. From there, participants will have the chance to react and share further reflections. In the third round, Google's AI will create declarative statements based on an analysis of all the previous responses. Finally, participants will vote on whether they agree or disagree with those statements, so they can see where they find common ground with others. The organizers are aware of how AI could show bias in its prompts or analysis of responses, or gloss over the nuances of very sensitive and personal topics as it tries to distill thousands of responses into broader themes. To help prevent that, Green said, the company will regularly evaluate the AI, and commit to transparency in making every participant's responses available in full. After the project concludes, Google will release a summary report of its findings and open source all of the responses, questions and prompts used. The company will also plan to brief policymakers, think tanks and academics on the results. For now, the effort is in the pilot stage, with hundreds of participants instead of thousands, and the project will officially kick off in September. Several details are either still being determined or haven't been released, like the specific topics that respondents will discuss. Respondents will be recruited by RepData, a third party vendor used by the Napolitan Institute. Google said it will share more details of its finalized recruiting approach closer to launch, but said RepData 'will follow a rigorous protocol in-line with industry standards.' Typically, pollsters use a method called 'random sampling,' which gives everyone in a certain population an equal chance of being included. Participants' voices are then sometimes 'weighted,' to account for populations that are less likely to participate in polls (like, for example, people without college degrees). For the We The People project, the goal is collecting a 'representative' sample of people for each congressional district. So, if a district is 90% white and 80% Republican, a representative set could reflect that, Rasmussen said. Green is optimistic about the results to come. 'I'm confident it's going to run like the conversation that you'd want if you were getting all of America in one place,' she said. More from Forbes Forbes Zuckerberg Squandered His AI Talent. Now He's Spending Billions To Replace It. By Rashi Shrivastava Forbes 'It Was Very Sudden': ChatGPT Users Mourned The Loss Of GPT-5's Predecessor By Richard Nieva Forbes You're Not Imagining It: AI Is Already Taking Tech Jobs By Richard Nieva
