Brief IA

Instagram and Mosseri: Controversial AI Cost-Cutting

🤖 Models & LLM·Tom Levy·

Instagram and Mosseri: Controversial AI Cost-Cutting

Instagram and Mosseri: Controversial AI Cost-Cutting
Key Takeaways
1Adam Mosseri announces a reduction in AI spending at Instagram, mentioning the halt of projects deemed absurd.
2The CEO criticizes the practice of tokenmaxxing, which has been adopted by Meta but rejected by Silicon Valley.
3Mosseri anticipates a decrease in AI costs, despite increased token usage by employees.
💡Why it mattersManaging AI costs is crucial for the competitiveness of tech companies in light of rising token usage and prices.
Le brief IA que lisent les pros

Le brief IA que les pros lisent chaque soir

Les 7 actus IA du jour, décryptées en 5 min. Gratuit.

Inclus dès l'inscription : notre sélection des meilleurs guides & comparatifs IA.

Choisis ton rythme

Gratuit · Pas de spam · Désabonnement en 1 clic

📄
Full Analysis

Instagram Reduces AI Spending Under Mosseri's Leadership

Instagram's CEO, Adam Mosseri, recently revealed that the platform has begun to cut its spending related to artificial intelligence. This decision was driven by the elimination of what he termed "absurd things" that the company had implemented. Mosseri clarified that these adjustments were relatively straightforward to make.

In a discussion on "Lenny's Podcast," Mosseri explained that cost reduction was facilitated by the removal of projects he deemed unnecessary. Although he did not detail these initiatives, he emphasized how easily resources can be wasted on projects that consume tokens without providing real added value.

The Challenges of Tokenmaxxing and Meta's Position

Mosseri also addressed the issue of tokenmaxxing, a practice that emerged in Silicon Valley, which involves maximizing the use of tokens to power AI models. Meta, Instagram's parent company, was one of the firms to adopt this approach, but it has since faced significant criticism and has been abandoned by many tech companies.

Tokens are essential data units for the functioning of large language models, such as those used by OpenAI's ChatGPT. Most AI providers charge for their services based on the consumption of these tokens, which can quickly escalate costs.

The Rising Costs of AI and Future Outlook

Mosseri's comments come at a time when many companies, both large and small, are facing rising costs associated with AI. This increase is partly due to the growing adoption of AI by employees and the heightened complexity of tasks they are asking these tools to perform. Andrew Macdonald, Uber's Chief Operating Officer, recently expressed similar concerns regarding Uber's AI spending, which has not always yielded the expected results.

Mosseri remains optimistic about a potential decrease in AI costs. He anticipates that increased competition among companies for market share will lead to lower prices. However, he acknowledges that spending in his division may continue to rise in the short term due to the increased use of tokens by employees.

Integrating Tokens into Budget Management

Mosseri indicated that AI tokens are now integrated into Instagram's budgeting process, alongside other technological resources such as GPUs, storage, and RAM. He suggested that, in the near future, a single engineer's token consumption could represent a cost equivalent to their salary.

He also mentioned the possibility of introducing caps on token usage, proportional to the company's confidence in employees' ability to use them effectively. For now, Instagram has not yet implemented such caps, but Mosseri believes this may become necessary.

Restructuring Teams at Instagram

Alongside these budget adjustments, Instagram has also modified the structure of its teams. Mosseri explained that the average team size has been reduced from about a dozen members to around six or seven. This reduction aligns with a broader trend observed in the tech industry, where companies are seeking to enhance efficiency by reducing the number of collaborators to coordinate.

Previously, a typical team at Instagram included several engineers specialized in different areas, along with project managers and data scientists. Today, teams, or "pods," consist of four to six more generalist engineers and one product staff member, often a specialist tailored to the specific needs of the project.

Mosseri concluded that this new structure allows for faster and more effective decision-making by reducing functional overlaps and adapting to industry changes.

Brief IA — L'actualité IA en français

L'essentiel de l'actualité de l'intelligence artificielle, décrypté et expliqué chaque jour.