A former OpenAI researcher has launched WaveForms AI, a startup focused on creating emotionally engaging voice interactions using AI. Backed by Andreessen Horowitz, the company has raised $40 million and is now valued at $200 million, according to CEO Alexis Conneau.
Conneau, known for his contributions to OpenAI’s GPT-4o voice mode, played a key role in advancing realistic AI conversations. The technology impressed users with its real-time responses and ability to handle interruptions but became embroiled in controversy. OpenAI was accused of copying Scarlett Johansson’s voice from the film Her, which the company denied, stating a different voice actress was used.
WaveForms AI, co-founded by Conneau and Coralie Lemaitre, plans to use its funding to develop advanced audio AI models. These models aim to make voice interactions with AI feel indistinguishable from human conversation, focusing on emotional connections. Conneau highlighted the importance of audio for delivering human-like, empathetic responses in AI systems.
The startup envisions creating consumer products that redefine human-computer interaction. Conneau said the goal is to provide immersive, enjoyable experiences that prioritise the quality of interaction over superintelligence. Further product details remain under wraps.
OpenAI has launched its text-to-video AI model, Sora, to ChatGPT Plus and Pro users, signalling a broader push into multimodal AI technologies. Initially limited to safety testers, Sora is now available as Sora Turbo at no additional cost, allowing users to create videos up to 20 seconds long in various resolutions and aspect ratios.
The move positions OpenAI to compete with similar tools from Meta, Google, and Stability AI. While the model is accessible in most regions, it remains unavailable in EU countries, the UK, and Switzerland due to regulatory considerations. OpenAI plans to introduce tailored pricing options for Sora next year.
The company emphasised safeguards against misuse, such as blocking harmful content like child exploitation and deepfake abuse. It also plans to gradually expand features, including uploads of people, as it enhances protections. Sora marks another step in OpenAI’s efforts to innovate responsibly in the AI space.
OpenAI is exploring the removal of a clause that restricts Microsoft’s access to its most advanced AI technology Artificial General Intelligence (AGI) once it is achieved. AGI, defined as a system that surpasses human capability in economically valuable tasks, has been excluded from Microsoft’s agreements under existing terms. The Financial Times reports that OpenAI aims to unlock further investments by lifting this restriction.
The clause, designed to safeguard AGI from misuse, currently gives OpenAI’s non-profit board control over such breakthroughs. Discussions within the board are ongoing, and no decision has been finalised. If the change proceeds, Microsoft could retain full access to future OpenAI advancements, even post-AGI, aligning with their significant backing of OpenAI.
This potential shift follows OpenAI’s restructuring efforts, including becoming a for-profit benefit corporation. In October, the company closed a $6.6B funding round, valuing it at $157B, as it continues redefining the AI market.
OpenAI is working to integrate AI into e-learning through customisable GPT tools, potentially revolutionising how students interact with academic content. According to Siya Raj Purohit of OpenAI‘s education team, professors are already using AI to create tailored course models, allowing students to engage with focused material. These tools could become staples in education, enabling personalised, lifelong learning.
The initiative complements OpenAI’s broader push into education, marked by the launch of ChatGPT Edu for universities and the hiring of former Coursera executive Leah Belsky. Despite these efforts, challenges remain as many educators express reservations about AI’s role in teaching. Tools like Khanmigo, developed with OpenAI, demonstrate AI’s potential but also reveal its current limitations, including accuracy issues.
With the education AI market expected to reach $88.2 billion, OpenAI is committed to refining its tools and addressing educators’ concerns to drive adoption in this burgeoning sector.
OpenAI has launched ChatGPT Pro, a premium version of its chatbot priced at $200 per month, designed for engineering and research applications. The new subscription tier joins the company’s existing lineup, including ChatGPT Plus, Team, and Enterprise, signalling OpenAI’s drive to broaden its technology’s industry adoption.
ChatGPT Pro offers users cutting-edge tools like OpenAI’s o1 reasoning model, o1 mini, GPT-4o, and an advanced voice interface. It also includes o1 pro mode, a high-performance feature utilising extra computational power to tackle complex queries, making it especially valuable for technical professionals.
According to OpenAI, o1 pro mode significantly outperforms earlier iterations like o1 and o1-preview on benchmarks for tasks in mathematics, science, and coding. This launch underscores OpenAI’s commitment to delivering sophisticated AI solutions tailored to specialised and demanding fields.
Anduril Industries and OpenAI have announced a partnership to advance AI applications for US national security. The collaboration will focus on enhancing counter-unmanned aircraft systems (CUAS), crucial for detecting and neutralising airborne drone threats.
By leveraging Anduril’s extensive CUAS data, AI models will be trained to respond to aerial threats in real time. OpenAI’s CEO, Sam Altman, highlighted the goal of safeguarding military personnel through these advanced AI solutions.
This partnership reflects the escalating global competition in AI-powered autonomous defence technologies, as nations like the United States and China race to innovate in automated military systems. Founded in 2017, Anduril specialises in autonomous systems, including drones and other tactical assets.
OpenAI is exploring advertising as a potential revenue model, according to a Financial Times report. While CFO Sarah Friar emphasised that there are ‘no active plans to pursue advertising,’ the company recently hired Shivakumar Venkataraman, a former Google ad executive, signalling interest in the possibility. OpenAI currently relies on subscriptions to fund its costly generative AI models, but rising expenses may prompt a shift in strategy.
The idea of ads doesn’t sit comfortably with OpenAI CEO Sam Altman, who described advertising as a “last resort” during a Harvard Business School chat. Altman has expressed unease about combining ads with AI, calling the prospect ‘uniquely unsettling.’ Still, the financial realities of sustaining and expanding AI tools may leave OpenAI with few alternatives.
As OpenAI balances innovation with business pressures, any move toward advertising could reshape how users engage with tools like ChatGPT. For now, the debate underscores the tension between maintaining accessibility and meeting operational demands.
Five Canadian news companies have launched a lawsuit against OpenAI, claiming its AI systems violate copyright laws. Torstar, Postmedia, The Globe and Mail, The Canadian Press, and CBC/Radio-Canada allege the company uses their journalism without permission or compensation. The legal filing, made in Ontario’s superior court, seeks damages and a permanent ban on OpenAI using their materials unlawfully.
The companies argue that OpenAI has deliberately appropriated their intellectual property for commercial purposes. In their statement, they emphasised the public value of journalism and condemned OpenAI’s actions as illegal. OpenAI, however, defended its practices, stating that its models rely on publicly available data and comply with fair use and copyright principles. The firm also noted its efforts to collaborate with publishers and provide mechanisms for opting out.
The case follows a trend of lawsuits by various creators, including authors and artists, against AI companies over the use of copyrighted content, and the Canadian lawsuit does not name Microsoft, a major OpenAI backer. Separately, Elon Musk recently expanded a legal case accusing both companies of attempting to dominate the generative AI market unlawfully.
Italy’s data protection authority has issued a warning to publisher GEDI over sharing personal data with OpenAI, citing potential violations of EU privacy regulations. GEDI, part of the Agnelli family’s Exor group, entered into a strategic partnership with OpenAI in September to provide Italian-language content for ChatGPT users.
Under the deal, OpenAI’schatbot would feature GEDI’s attributed content and links, while its journalism could refine the AI’s accuracy. Concerns have arisen due to the sensitive nature of the archives, containing information on millions of individuals. The regulator highlighted that such data requires careful handling and warned of potential sanctions if EU rules are breached.
GEDI clarified that the partnership does not involve selling personal data and noted that the project is still under review. No editorial content has been shared with OpenAI to date, according to a company statement. Discussions with the watchdog are ongoing, with GEDI expressing hope for constructive dialogue to resolve concerns.
Representatives from OpenAI have not yet commented on the matter.
Alibaba has unveiled QwQ-32B-Preview, a new reasoning AI model designed to rival OpenAI’s o1 series. With 32.5 billion parameters and support for prompts up to 32,000 words, the model surpasses competitors in specific benchmarks, including logic puzzles and maths tests. Available for download under a permissive Apache 2.0 licence, it introduces robust reasoning capabilities but also exhibits limitations like language switching and occasional lapses in common sense.
The model incorporates test-time compute, enabling more thorough problem-solving by planning its steps before providing answers. However, such reasoning processes may result in slower responses. Like other AI models made by companies in China, QwQ-32B complies with local regulatory requirements, including constraints on politically sensitive topics, reflecting national ideological alignment.
Reasoning models like QwQ-32B mark a shift in AI development as traditional scaling laws show diminishing returns. Major firms, including Google, are exploring similar approaches, highlighting the race to innovate AI capabilities globally.
