Reinforcement Learning from Human Feedback, or RLHF, is a method for shaping model behavior with human preferences. Reviewers compare or score candidate responses, those judgments train a reward model, and reinforcement learning then adjusts the language model to favor outputs that receive higher predicted rewards. RLHF can improve instruction following, helpfulness, tone, and safety beyond basic pretraining. Its results depend heavily on who provides feedback, how instructions are written, and whether the examples represent real users and edge cases. The process may reward superficial agreement or hide uncertainty, so it is commonly combined with automated evaluations, red teaming, policy rules, and ongoing post-deployment monitoring.
Goldman Sachs is projecting SpaceX’s AI revenues will grow roughly 100 times by 2030 – from $3.2 billion to $322 billion annually – figures the bank is using to anchor a $1.78 trillion IPO valuation for the company it is simultaneously leading to market.
Anthropic has called on the global AI industry to consider slowing or temporarily pausing frontier model development, warning that AI systems are already automating parts of their own creation and that full recursive self-improvement – where AI designs and trains its own successors without human involvement – could arrive within one to two years.
Microsoft has introduced seven proprietary AI models spanning reasoning, coding, image generation, transcription, and voice synthesis.
Meta has eased plans to track employee computer activity for AI training after internal criticism, introducing new controls that allow workers to pause data collection or request exemptions.
Anthropic has confidentially filed for an initial public offering, positioning itself for one of the most closely watched AI listings in history.
OpenAI has introduced role-specific plugins, collaborative websites, and annotation tools for Codex as the platform expands beyond software development. More than 5 million people now use Codex weekly, with non-developers becoming one of its fastest-growing user groups.
Investors and technology leaders increasingly view humanoid robots as the next major growth opportunity in artificial intelligence. Industry forecasts suggest the market could expand from a few billion dollars today to hundreds of billions within the next decade.
Anthropic says cybercriminals are increasingly using AI for advanced post-compromise operations and autonomous attack chains. The company argues that existing cybersecurity frameworks may no longer fully capture how AI-powered threats operate.
Anthropic is expanding Project Glasswing to approximately 200 organizations using Claude Mythos Preview to identify software vulnerabilities. The initiative aims to strengthen cybersecurity defenses as increasingly powerful AI models become widely available.
Morningstar has valued SpaceX at $780 billion, far below the company’s reported $1.75 trillion IPO target. The research firm cited uncertainty around SpaceX’s AI business, Starlink’s long-term prospects, and unproven technologies.