The week in AI: Roadmap to GPT-5, Musk's OpenAI takeover bid, Vance's bold & brash speech at Paris Summit

In partnership with

Welcome to The Dispatch! We are the newsletter that keeps you informed about AI. Each Thursday, we aggregate the major developments in artificial intelligence - we pass along the news, useful resources, tools and services; we highlight the top research in the field as well as exciting developments in open source. Even if you aren’t a machine learning engineer, we’ll keep you in touch with the most important developments in AI.

NEWS & OPINION

The roadmap to GPT-5: a “unified intelligence” model

-------------------------

Wednesday on X/Twitter, Sam Altman himself laid out OpenAI’s roadmap towards GPT-5, while also teasing an interim release of GPT-4.5. Here’s what to know:

• The tentative timeline for GPT-4.5 is weeks, and months for GPT-5. We’re expecting a late April or May release for GPT-5. GPT-4.5 (internally called Orion) will be the last ‘non-reasoning’/Chain of Thought model from OpenAI. CoT has been shown to improve LLMs’ output quality significantly, particularly when they tackle complex reasoning tasks, but at significantly higher inference costs. GPT-4.5 will likely be the “last, best” foundation model that doesn’t use the secret sauce powering reasoning models like DeepSeek’s R1 and OpenAI’s o1.

• Altman has long admitted that his company needs a naming scheme revamp, and in his post he claims the company wants to get away from their hard to follow product/model offerings and move towards a unified intelligence model (GPT-5) that “just works”. So, GPT-5 will be a system with different levels of intelligence based on pricing tiers and user needs. It’ll have all of ChatGPT’s (currently quite disparate) tools like Search, Canvas, Thinking, Deep Research etc. directly integrated. Users will no longer have to constantly pick from a dozen different options and models based on use cases.

• That sounds great - assuming users will still have the ability to manually choose, if they want - but there’s likely another angle at work for OpenAI here. In the post, Altman says they will no longer be shipping o3 (their next-gen advanced reasoning model) as a standalone product. Sam has previously not been shy about admitting to losing money on the $200/mo ChatGPT Pro subscription tier due to higher-than-expected use rates of their most expensive to run models, like o1. Given o3’s absolutely bonkers cost to run, you could argue GPT-5 is as much about a unified way for the company to have some default-level control over its surging compute requirements, which not even Microsoft could cater to, as it is about “unified intelligence”.

Additionally, it looks like OpenAI is trying to cast a very wide net to lure in more free users: GPT-5 at standard intelligence will be available for all free users with unlimited use (barring abuse). We’re guessing that means unlimited GPT-4.5 model use, with limited daily/weekly access to the more advanced models/features, similar to their current structure for the free tier. Working in that direction: soon, Deep Research will be available to Plus (10 queries/mo) and free ChatGPT users (2 queries/mo). o3-mini can also now see images in ChatGPT and the limit for o3-mini-high is now 50 messages/day (up from 50/week on Plus tier).

Altman made a lengthy personal blog post that covers three observations he believes to be true about the development of AI towards AGI, and the impacts on society to follow.

Elon Musk-led group’s $97 billion bid for control of OpenAI: what to know

-------------------------

As if Elon Musk wasn’t busy enough this week, he also found the time to spearhead a consortium of investors’ unsolicited attempt to purchase OpenAI for $97.4 billion. The offer was backed by Musk's own company, xAI, in addition to several investors in Musk's other businesses.

Altman, responding from the Paris AI Summit, shifted the offer's decimal point in a snarky response on X, saying, "no thank you but we will buy twitter for $9.74 billion if you want." He then made several media appearances clarifying that OpenAI was not for sale, and when asked by Bloomberg whether he thought Musk’s offer came from a position of insecurity, Altman responded, “Probably his whole life is from a position of insecurity, I feel for the guy... I don't think he's, like, a happy person.”

Beef between Musk and Altman goes back to 2015, when the pair partnered (with others) to co-found OpenAI as a nonprofit. Musk cut ties with the company in 2018 but watched from the sidelines as OpenAI became a media darling in late 2022 following the public launch of ChatGPT.

Beyond the drama, the bid does have major implications for OpenAI. Musk's move isn't just for show, even if he has zero intention to purchase; it directly impacts OpenAI's ongoing and crucial effort to transition from a non-profit to a for-profit entity. The conversion is necessary for OpenAI to secure more funding and simplify its structure, but it requires valuing and compensating the original non-profit arm that’s still in control. Musk's enormous $97 billion offer publicly establishes a floor price for this non-profit stake. Suddenly, whatever internal valuations Altman was working with are now overshadowed by this very public, very large number.

This bid also significantly raises the stakes for Altman's for-profit conversion plan. Any deal he strikes to compensate the non-profit will now be scrutinized against Musk's offer – not just by OpenAI's board, but also potentially by regulators. To gain approval, Altman's plan will need to be perceived as equally or even more beneficial than Musk's proposal. Essentially, Musk has strategically inserted himself into OpenAI's financial restructuring, forcing Altman to potentially pay more to complete the for-profit transition and adding a layer of complexity and public pressure to an already delicate process.

The Paris AI Summit happened

-------------------------

Many haven't realized it yet, but the global AI governance landscape is shifting dramatically and rapidly. This week's AI Summit in Paris revealed two main narratives:

• The EU wants to signal that it's ready for the AI race and is willing to do whatever it takes to stay competitive, particularly against China and the US. It even went so far as to withdraw the AI Liability Directive. Considering how fervently the EU has pushed a heavy regulation narrative, there’s a bit of a “reeds blowing in the wind” feeling here. French President Emmanuel Macron said: "Europe is going to accelerate; France is going to accelerate - and so for us, France, we're announcing tomorrow at this summit €109 billion of investment in AI over the next few years."

• The US, now under the new administration, is advancing a full-blown deregulation strategy. It has revoked Biden’s Executive Order on AI and made it clear this week that it will not accept regulatory constraints from the EU. The UK and US also refused to sign international AI declaration. Ouch. Vice President JD Vance stated during a particularly fiery speech about US AI sovereignty: “The Trump administration is troubled by reports that some foreign governments are considering tightening the screws on US tech companies with international footprints. Now America cannot and will not accept that.”

It's no longer 2024, and 2025 brought Trump and a major narrative shift that includes deregulation. The EU is openly changing its narrative to reposition itself as a more competitive player. Recent EU reports, such as this one by former European Central Bank President Mario Draghi, have intensified the internal pressure.

Given the EU AI Act’s numerous loopholes and exceptions - combined with the EU’s digital chief Henna Virkkunen's declaration to the audience that the rules will be applied in a “business-friendly” way - it’s pretty clear what was once seen as a potential guidepost for regulation has already lost most its teeth.

MORE IN AI THIS WEEK

• Anthropic publishes first Economic Index - a new study tracking AI’s labor market impact

• Thomson Reuters wins first major AI copyright case in US

• I let ChatGPT’s new ‘agent’ manage my life. It spent $31 on a dozen eggs

• House lawmakers push to ban Chinese AI app DeepSeek from US government devices

• AI chatbots unable to accurately summarize news, BBC finds

• ChatGPT may not be as power-hungry as once assumed

• Scarlett Johansson calls for deepfake ban after anti-Kanye AI video goes viral

• Over 5,000 artists signed an open letter calling for the cancellation of ‘Augmented Intelligence,’ an upcoming AI art auction in NY, arguing AI exploits and devalues human artists

• Apple partners with Alibaba to bring iPhone AI features to China

• OpenAI will reportedly finalize the design for its first generation of in-house AI chips, will work with TSMC on the initial fabrication

Hire Ava, the Industry-Leading AI BDR

Your BDR team is wasting time on things AI can automate. Our AI BDR Ava automates your entire outbound demand generation so you can get leads delivered to your inbox on autopilot.

She operates within the Artisan platform, which consolidates every tool you need for outbound:

• 300M+ High-Quality B2B Prospects, including E-Commerce and Local Business Leads

• Automated Lead Enrichment With 10+ Data Sources

• Full Email Deliverability Management

• Multi-Channel Outreach Across Email & LinkedIn

• Human-Level Personalization

Book a demo to see what Ava can do.

TECHNICAL NEWS, DEVELOPMENT, RESEARCH & OPEN SOURCE

• ChatGPT “turning right”? An experimental study on the political value shift in large language models

• DeepScaleR: Surpassing o1-preview with a 1.5B model by scaling reinforcement learning

• TWO AI introduces SUTRA-R0: a multilingual-focused reasoning model that surpasses DeepSeek-R1 in Indian, Korean, Japanese language benchmarks

• Gold-medalist performance in solving Olympiad geometry with Google DeepMind’s AlphaGeometry2

• Harvard releases 311k datasets from Data.gov (16TB), updated daily for open access

• Hugging Face publishes the first open source robotics foundation models for real-world applications

• s1-32B: improves test-time scaling with budget forcing, boosting math reasoning and competition scores

That’s all for this week! We’ll see you next Thursday.