Uber president says AI spending is getting 'harder to justify'
Tech workers are questioning whether massive corporate spending on AI—especially “tokenmaxxing” with large language models—delivers real business value. Commenters point to Uber’s rapid exhaustion of its annual AI budget, marginal end‑user feature gains, and the difficulty of tying AI tooling to measurable productivity or profit, particularly in large organizations where code production is rarely the main bottleneck. Many see AI’s current impact as a mix of localized wins, hype‑driven FOMO, and management attempts to force adoption, with growing skepticism about the sustainability of current cost levels.
Perceived Value of Uber’s AI Spend
- Many commenters say the Uber rider experience feels unchanged; they question where all the “billions of tokens” went.
- Examples from the article (hotel bookings, travel recommendations, in-ride food pickup, voice bookings) are widely seen as marginal or gimmicky, with little core value.
- Some argue that invisible backend work (regulatory compliance, payments, reliability, food delivery) already justified traditional engineering costs, but similar justification for AI spend is missing.
Tokenmaxxing, Metrics, and Management Fads
- Strong criticism of “tokenmaxxing” (measuring success by tokens burned) as a textbook case of Goodhart’s law: people optimize the metric instead of outcomes.
- Several report companies forcing AI adoption across all staff, even while skimping on proven tools, because investors and executives demand “AI usage.”
- This behavior is compared to past tech fads (SOA, Hadoop, Kubernetes, cloud, etc.), but AI is seen as unusually global and cross‑industry in its pressure and FOMO.
Productivity, Code Quality, and Bottlenecks
- Some engineers find real gains for specific tasks (e.g., test generation, UI automation, internal tools) but note these often don’t show up directly in quarterly results.
- Others say non‑coding bottlenecks (process, review, security, org politics) dominate in large orgs, so faster coding doesn’t translate into business impact.
- Concerns include: reduced understanding of code, increasing technical debt, and the risk of “90% correct” AI features breaking large, critical systems.
Strategic Motives and Labor Dynamics
- One view: massive AI spend is a strategic race to discover how to truly automate software development and gain a generational advantage.
- Another view: it’s partly about weakening software engineers’ bargaining power by signaling they are replaceable, even if current tools don’t actually achieve that.
- Some expect AI investments to be followed by workforce reductions and rising expectations for remaining staff, while token budgets get cut.
Bubble, Sustainability, and Future Direction
- Many doubt the economics: modest productivity gains versus huge token bills and infrastructure costs.
- Some foresee an AI hype correction, especially outside tech, where firms are already scaling back due to lack of measurable ROI.
- A minority believes the long‑term “way of the future” is smaller, domain‑specific or local models, but timing and viability are seen as unclear.