\n<\/p>\n
To cap off a day of product releases<\/a>, OpenAI researchers, engineers, and executives, including OpenAI CEO Sam Altman, answered questions in a wide-ranging Reddit AMA<\/a> on Friday.<\/p>\n OpenAI the company finds itself in a bit of a precarious position. It\u2019s battling the perception that it\u2019s ceding ground in the AI race to Chinese companies like DeepSeek<\/a>, which OpenAI alleges might\u2019ve stolen its IP. The ChatGPT maker has been trying to shore up its relationship with Washington<\/a> and simultaneously pursue an ambitious data center project<\/a>, while reportedly laying groundwork<\/a> for one of the largest financing rounds in history.<\/p>\n Altman admitted that DeepSeek has lessened OpenAI\u2019s lead in AI, and he also said he believes OpenAI has been \u201con the wrong side of history\u201d when it comes to open-sourcing its technologies. While OpenAI has open-sourced models in the past, the company has generally favored a proprietary, closed-source development approach.<\/p>\n \u201c[I personally think we need to] figure out a different open source strategy,\u201d Altman said. \u201cNot everyone at OpenAI shares this view, and it\u2019s also not our current highest priority [\u2026] We will produce better models [going forward], but we will maintain less of a lead than we did in previous years.\u201d<\/p>\n In a follow-up reply, Kevin Weil, OpenAI\u2019s chief product officer, said that OpenAI is considering open-sourcing older models that aren\u2019t state-of-the-art anymore. \u201cWe\u2019ll definitely think about doing more of this,\u201d he said, without going into greater detail.<\/p>\n Beyond prompting OpenAI to reconsider its release philosophy, Altman said that DeepSeek has pushed the company to potentially reveal more about how its so-called reasoning models, like the o3-mini model released today, show their \u201cthought process.\u201d Currently, OpenAI\u2019s models conceal their reasoning, a strategy intended to prevent competitors from scraping training data for their own models. In contrast, DeepSeek\u2019s reasoning model, R1<\/a>, shows its full chain of thought.<\/p>\n \u201cWe\u2019re working on showing a bunch more than we show today \u2014 [showing the model thought process] will be very very soon,\u201d Weil added. \u201cTBD on all \u2014 showing all chain of thought leads to competitive distillation, but we also know people (at least power users) want it, so we\u2019ll find the right way to balance it.\u201d<\/p>\n Altman and Weil attempted to dispel rumors<\/a> that ChatGPT, the chatbot platform through which OpenAI launches many of its models, would increase in price in the future. Altman said that he\u2019d like to make ChatGPT \u201ccheaper\u201d over time, if feasible.<\/p>\n Altman previously said that OpenAI was losing money on its priciest ChatGPT plan<\/a>, ChatGPT Pro, which costs $200 per month.<\/p>\n In a somewhat related thread, Weil said that OpenAI continues to see evidence that more compute power leads to \u201cbetter\u201d and more performant models. That\u2019s in large part what\u2019s necessitating projects such as Stargate, OpenAI\u2019s recently announced<\/a> massive data center project, Weil said. Serving a growing user base is fueling compute demand within OpenAI, as well, he continued.<\/p>\n Asked about recursive self-improvement that might be enabled by these powerful models, Altman said he thinks a \u201cfast takeoff\u201d is more plausible than he once believed. Recursive self-improvement is a process where an AI system could improve its own intelligence and capabilities without human input.<\/p>\n Of course, it\u2019s worth noting that Altman is notorious for overpromising. It wasn\u2019t long ago that he lowered OpenAI\u2019s bar for AGI<\/a>.<\/p>\n One Reddit user asked whether OpenAI\u2019s models, self-improving or not, would be used to develop destructive weapons \u2014 specifically nuclear weapons. This week, OpenAI announced a partnership with the U.S. government<\/a> to give its models to the U.S. National Laboratories in part for nuclear defense research.<\/p>\n Weil said he trusted the government.<\/p>\n \u201cI\u2019ve gotten to know these scientists and they are AI experts in addition to world class researchers,\u201d he said. \u201cThey understand the power and the limits of the models, and I don\u2019t think there\u2019s any chance they just YOLO some model output into a nuclear calculation. They\u2019re smart and evidence-based and they do a lot of experimentation and data work to validate all their work.\u201d<\/p>\n The OpenAI team was asked several questions of a more technical nature, like when OpenAI\u2019s next reasoning model, o3<\/a>, will be released (\u201cmore than a few weeks, less than a few months,\u201d Altman said); when the company\u2019s next flagship \u201cnon-reasoning\u201d model, GPT-5<\/a>, might land (\u201cdon\u2019t have a timeline yet,\u201d said Altman); and when OpenAI might unveil a successor to DALL-E 3, the company\u2019s image-generating model<\/a>. DALL-E 3, which was released around two years ago, has gotten rather long in the tooth. Image generation tech has improved by leaps and bounds since DALL-E 3\u2019s debut, and the model is no longer competitive on a number of benchmark tests<\/a>.<\/p>\n \u201cYes! We\u2019re working on it,\u201d Weil said of a DALL-E 3 follow-up. \u201cAnd I think it\u2019s going to be worth the wait.\u201d<\/p>\n<\/div>\n