Dr Doolittle eat your heart out, fans of the iconic 90’s cult classic will be in awe of the latest project developing at Google Deepmind, namely DolphinGemma. On National Dolphin Day, in collaboration with Georgia Tech and the Wild Dolphin Project, Google announced their latest AI foundational model which deciphers the mysterious clicks of the second most intelligent creatures on the planet. The model helps researchers discover patterns, structure, and eventually predict what dolphins are going to do next, much like with what a Large Language Model (LLM) does with human language.
Beginning 1 September 2025, China will mandate artificial intelligence education for all primary and secondary school students, including children as young as six. The curriculum is designed to be age-appropriate: younger students will engage in interactive activities introducing basic AI concepts, while older students will delve into more advanced topics such as machine learning, robotics, and real-world AI applications. Schools have the flexibility to integrate AI education into existing subjects like science and technology or offer it as standalone courses.
With China already leading in AI development, this initiative ensures that future generations will grow up ‘AI-literate’. It feels like another iPhone moment – where a shift today reshapes the world tomorrow.
In a recent interview with Time magazine, DeepMind CEO Demis Hassabis reflected on the high-stakes future of AI, even as he celebrated winning the 2024 Nobel Prize in Chemistry. While optimistic about Artificial General Intelligence’s (AGI) potential to cure disease and solve climate change, Hassabis remains deeply concerned about its risks. “I’d be very worried about society today if I didn’t know that something as transformative as AI was coming down the line”, he told Time.
That is the issue, people in everyday life do not realise the shift, this technology is going to have. The AI revolution is expected to have an impact on society as profound and far-reaching as the industrial revolution. Artificial General Intelligence is reachable and all industry leaders/governments goal is to achieve this. People need to become more aware of what is coming in the next five to ten years, whether the outcome will be positive or negative remains to be seen. What is clear is though, is that the developments in science and technology will attain an unimaginable standard.
Here a more positive snippet of the interview:
A Swiss-led study presented at the ESCMID Global Congress unveiled a significant advancement in tuberculosis (TB) diagnostics. The research introduced ULTR-AI (max), an AI-driven lung ultrasound interpretation system, which achieved a sensitivity of 93% and specificity of 81% in detecting pulmonary TB. These metrics surpass the World Health Organization’s (WHO) minimum targets to achieve the gold standard.
The Swaasa AI platform, developed by Salcit Technologies, offers a non-invasive approach to TB screening by analysing cough sounds via smartphones. While specific accuracy figures for TB detection were not detailed in the available sources, Swaasa has shown 90% accuracy in distinguishing between healthy individuals and those with respiratory conditions, as well as differentiating normal from abnormal spirometry results.
Lots of long words, but the progress AI is making on the medical front is unquestionable, with Bill Gates saying last month that AI will replace teachers and doctors in 10 years.
As is often the case in these AI reviews, here is a published paper that has not yet been peer-reviewed but has some interesting findings. The Turing Test is a measure of a machine’s intelligence, evaluating whether it can exhibit human-like conversation indistinguishable from that of a real person. For the first time ever, AI has passed this test. If confirmed, this milestone would mark a historic turning point in the development of artificial intelligence and our understanding of machine cognition.
In the study conducted by researchers at the University of California, San Diego, OpenAI’s GPT-4.5 became the first AI model to convincingly pass a rigorous, three-party version of the Turing Test. Participants engaged in simultaneous five-minute text conversations with both GPT-4.5 and a human. When GPT-4.5 was prompted to adopt a human-like persona, it was judged to be the human 73% of the time – surpassing the actual human participants.
This month saw numerous new models released:
OpenAI introduced GPT-4.1, enhancing coding capabilities, instruction adherence, and long-context understanding. Complementing this, the o3 model excels in complex reasoning tasks across domains like math and science, while o4-mini offers efficient performance in coding and visual tasks.
Kling AI 2.0 continues to innovate in AI-driven video generation, enhancing content creation processes.
ByteDance’s Seaweed-7B is a 7-billion-parameter model trained for video generation, demonstrating competitive performance with efficient training.
Elsewhere Harvard released what the majority of people use AI for, and China wants to have flying taxis within three years.
The world is changing and it is imperative to stay up to date with all developments in the field, read the previous month’s AI reviews here.
