As AI voices become increasingly difficult to detect, ElevenLabs is incorporating Google's SynthID to assist you in identifying the fakes.

      Unsplash

      In the past, identifying AI-generated content was almost like a game. Images featured oddities like extra fingers, chatbots wrote in a manner reminiscent of overly eager interns, and synthetic voices had a distinctly robotic quality. Those times are quickly fading.

      Nowadays, AI voices can laugh, whisper, pause naturally, and even convey emotions convincingly, often tricking many listeners. While this is thrilling for creators, it also raises an increasing trust issue. If distinguishing between real and AI-generated clips becomes difficult, how can one discern what to trust?

      Concealing the truth for later revelation

      This is the issue that ElevenLabs is addressing. The company has announced its integration of Google DeepMind's SynthID watermarking technology into AI-generated speech, starting with text-to-speech audio produced by free users, with plans to expand this across all audio generations in the upcoming weeks.

      Google

      The concept is surprisingly intelligent. Instead of using metadata that can easily vanish when a file is edited or shared, SynthID embeds an inaudible digital watermark directly within the audio itself. Although you won’t hear it, ElevenLabs claims the watermark endures common edits such as trimming, compression, speed alterations, file conversions, and even the removal of metadata. Along with the watermarking rollout, ElevenLabs will also introduce a free Audio Detector, allowing anyone to determine if a recording was made using its platform.

      Trust may become AI’s most essential feature

      The timing is appropriate. AI-generated audio is increasingly challenging to differentiate from authentic recordings, and deepfake scams are becoming more advanced. While ElevenLabs already monitors content internally and forbids deceptive uses of its platform, incorporating a persistent watermark adds a level of accountability that remains intact even after a file has left its servers. Additionally, persistent watermarks could eventually assist creators in proving ownership of AI-generated work, maintaining content credentials, and simplifying tracking of copyrighted material on platforms like YouTube, Instagram, and TikTok.

      ElevenLabs

      Certainly, watermarking will not eliminate harmful deepfakes overnight. Malicious actors will continue seeking ways to evade detection systems. However, as AI-generated audio approaches indistinguishability from human speech, the underlying invisibility of its origins could become just as crucial as the technology that produces the voices initially. The future of AI is not solely focused on improving the human-like quality of synthetic voices; it may also hinge on ensuring that people are aware when those voices are not genuine.

      Shimul is a contributor at Digital Trends, with over five years of experience in the technology field.

      Stanford scientists developed an AI that can create healthier, greener burgers

      The new system balances nutrition, flavor, cost, and environmental effects to generate improved recipes.

      Artificial intelligence has already been utilized to write code, discover medications, and produce videos. Now, it’s aiming to craft a superior burger. Researchers from Stanford University have introduced BurgerAI, a novel AI system that designs burger recipes while considering taste, nutrition, sustainability, and cost. Interestingly, in blind taste tests, some diners enjoyed the AI-generated burgers just as much or even more than a well-known fast-food burger.

      BurgerAI is intended to generate original recipes, not replicate existing ones.

      OpenAI unveils its most advanced GPT-5.6 model, but access is currently restricted

      GPT-5.6 offers enhanced reasoning, autonomy, and cybersecurity features, but its rollout is presently limited to government-approved clients.

      OpenAI has officially introduced GPT-5.6, its most advanced AI model series to date. However, there's a caveat: unless you're among a select few approved customers, you won’t be able to experiment with it anytime soon. Instead of a widespread launch, the company is initiating a tightly regulated preview while navigating a new U.S. government review procedure.

      GPT-5.6 is available, but only a limited audience can utilize it.

      Bacteria might hold the key to cooling your computers

      Researchers discover that bacteria can cultivate thermal materials that outperform current cooling technologies.

      PC and laptop temperatures are rising annually, especially as AI tools and resource-intensive software increasingly push them to their limits. Heat is a significant performance bottleneck, prompting users to implement extreme solutions like attaching actual ice machines to RTX GPUs. A recent study highlights a far less drastic solution to this issue, involving bacteria (via TechXplore).

      Researchers have devised a new method to cultivate thermal interface materials—the substances that sit between a chip and its cooling system to enhance heat dissipation. This process involves feeding bacteria sugar and metal ions, allowing the microbes to naturally produce the material rather than relying on conventional chemical manufacturing.