• “We aim only to have high performers at Netflix—people who are great at what they do, and even better at working together.” 
• The original manifesto was a public pushback on the notion that employers are families

• People Over Process: Empowerment Through Freedom
• In practice, this means empowering employees to understand what it takes to get the job done and deliver on that promise.
• This approach has been instrumental in maintaining agility and creativity within the company.

• Uncomfortably Exciting: Embracing Ambition and Risk
• This willingness to embrace risk and discomfort can also create internal conflict in this age of employee advocac
• Great and Always Better: A Commitment to Continuous Improvement
• “We often say Netflix sucks today compared to where we can be tomorrow.”
• By owning and proactively sharing their ways of working (selflessness, candor, being performance-driven), Netflix allows future employees to opt in or out based on self-alignment
• By maintaining a focus on excellence, collaboration, empowerment, ambition, and continuous improvement, Netflix aims to remain a place where talented individuals can do “the best work of their lives.”
• That’s where the partnership between strategic HR (or “Talent” in Netflix’s case) and business operations come into play

10 more annotations...

• A Gartner survey of 85 HR leaders conducted in February revealed that only 35% are confident their current approach to HR technology aligns with achieving business objectives
• highlights the importance of helping human resource leaders understand new tech’s ability to generate strategic business value.
• This view adds a layer of insight to the common belief that HR technology will allow people to focus on the individual tasks they enjoy most, a vantage point that Hudson says does not inherently benefit the organization or the employee.
• He says that plans to apply new technology should ensure employees are fully equipped to perform their jobs, not just eliminate tasks.
• Gartner’s research also highlights the organizational benefits of what they call an “augmented approach” to implementing HR tech. This perspective enhances HR by promoting the adoption of HR technology and leveraging it to generate high-impact contributions that maximize business value. The alternative is a “capacity-centric” approach, which emphasizes increased productivity or doing more with less.
• Gartner’s research found that two out of three HR leaders believe their function’s effectiveness will diminish without improvements in their approach to technology, leaving a gap between current practices and future potential.
• There may be several reasons for this. Technology has often automated tasks that HR professionals felt competent in, tasks that brought immediate positive feedback from their internal customers. This shift can increase the intensity of their roles, as they are left with more challenging or escalated responsibilities.
• . Self-service technologies and automation can reduce these personal interactions, thus diminishing opportunities for these valued moments.
• To bridge this gap, technology can help HR professionals by providing data to focus workplace efforts and identify underlying problems that might otherwise go unnoticed
• Hudson recommends creating technology plans or roadmaps that clearly show the connection between these smaller changes and the ultimate capabilities the organization aims to achieve. Breaking digital transformation projects into manageable segments
• For HR leaders aiming to present incremental plans rather than a large transformational goal, several strategies have proven effective, says Hudson
• Essentially, working back from the fully functioned, almost ‘imagined’ version,” according to Hudson. By breaking down the path to this advanced state into achievable milestones, he says, HR leaders can tell a compelling story and gain buy-in for initial changes.
• However, they may feel threatened by technology and therefore not look for new use cases for the technology that could allow it to deliver much more value, Hudson says. Gartner calls this a “preservation mindset.”
• Understanding what technology will remove from their roles makes HR staff protective of the areas where they currently add value.
• Because there is a tendency in HR functions to see the technology possibilities within the framework of the team’s corner of the department—recruiting, learning and development, benefits, etc.—human resources as a cohesive unit loses sight of the experience for the end user, who is usually a manager or an employee outside of the HR department.
• half of HR staff agree that HR technology solutions have damaged HR’s reputation across their organization, according to a 2024 Gartner study of over 600 practitioners.
• Hudson advises HR practice leaders to create shared goals for the success of the overall HR technology portfolio. In some very complex areas, this may require the creation of roles targeted at coordinating technologies as “products,” says Hudson, with a greater focus on what the end user,

16 more annotations...

• Des milliers de milliards de tokens -à savoir un mot ou partie de mot- sont utilisés pour entraîner le LLM de ChatGPT 4. L’IA générative de Meta, LLama 3 aurait nécessité 15.000 milliards de Tokens. Les ordres de grandeur du volume de données entrantes sont tout aussi colossaux pour les autres LLM telles Claude d’Antropic, Gemini de Google et autres IAGen.
• Principal résultat, si les tendances actuelles du développement des LLM se poursuivent, les modèles seront entraînés sur des ensembles de données d’une taille à peu près égale à la totalité du stock disponible de données publiques, entre 2026 et 2032.
• Face à la pénurie possible de données textuelles humaines publiques, les chercheurs de Cornell explorent les solutions pour contourner ce goulet d’étranglement. Trois pistes sont envisagées.
• D’une part, la génération de données synthétiques et non-publiques créées par des algorithmes qui reproduisent les propriétés statistiques des données réelles, sans révéler d’informations sur des personnes réelles pour préserver la confidentialité
• est à prévoir que cette solution sera très problématique à mettre en œuvre sur le plan législatif.
• D’autre part, l’étude envisage l'apprentissage à partir de l'utilisation de données non publiques. Là aussi, la voie parait impossible à suivre
• La troisième solution mentionnée par l’étude est plus crédible et réaliste, celle de l’optimisation des modèles de machine learning

6 more annotations...

• For HR leaders, this means ensuring AI tools used within their organizations comply with the directive’s requirements for data quality and bias mitigation.

   

  AI can detect and address biases in hiring practices, compensation structures, and performance evaluations, prompting human oversight to correct disparities.

• AI-powered recruitment tools can significantly enhance the diversity of candidate pools.

   

  The AI Act’s transparency requirements mean AI systems must clearly communicate their functioning, helping remove biases from the hiring process.

• The AI Act promotes the use of AI to foster inclusive communication.

   

  AI-driven tools for automatic translation, transcription, and summarization can break down language barriers and make communication more accessible to employees with disabilities.

• AI can provide personalized learning experiences that cater to diverse employee needs.

   

  AI-driven learning platforms can tailor training programs to individual learning styles and career development paths, creating a more inclusive environment where all employees have equal growth opportunities.

2 more annotations...

• Alors que 90 % des répondants l’utilisent pour divers usages, des réserves sont formulées sur la capacité de l’infrastructure à prendre en charge l’intelligence artificielle.
• A noter, plus d'un tiers des personnes interrogées (38 %) disent que leur entreprise utilisait déjà l'IA pour rendre les opérations plus efficaces. Avant donc l’irruption de ChatGPT dans le champ économique et médiatique en novembre 2022.
• Seuls 43 % des répondants affirment que les bases de données de leur entreprise peuvent répondre aux besoins accrus de l'IA.
• euls 38 % des participants à l’étude faisant très confiance à la qualité des données et à la formation utilisée dans les technologies d'IA. Ils jugent qu’il s’agit d’un obstacle majeur à l'adoption de l'IA.
• Hors de question de lui faire prendre des décisions, ce qui dérive du bon sens en l’état des capacités des IA. Les craintes en termes de confidentialité et de sécurité sont des obstacles à l'adoption des IA.
• Plus précisément, les expériences négatives sont citées par quatre participants sur dix (41 %). Un taux très important pas conséquent. En détail, les préoccupations relatives à la confidentialité (48 %) et les risques de sécurité (43 %) sont les plus souvent citées.

4 more annotations...

• Un gros fourre-tout que l’on peut difficilement qualifier de « marché » tant les offres disponibles et technologies utilisées sont différentes…
• le « marché » de l’IA générative est en réalité composé de services en ligne reposant sur des briques technologiques payantes ou gratuites, mais qu’il n’y a pas encore réellement de produits grand public capables de séduire des centaines de millions d’utilisateurs
• Donc pour le moment, le « marché » de l’IA générative se résume à une infinité de startups qui proposent toutes plus ou moins la même chose (des outils de génération de contenus) reposant sur des modèles payants édités par des sociétés privées de petites et grandes tailles (OpenAI, Anthropic, Mistral, Google, Microsoft, Amazon…).
• La stratégie d’Apple pour intégrer l’intelligence artificielle à ses terminaux et services ne déroge pas à la règle, car ils proposent, ou plutôt vont proposer une approche minimaliste, mais parfaitement maitrisée pour pouvoir garantir une expérience sans faille : Apple Intelligence.
• toutes les interactions se feront à travers des commandes simplifiées pré-définies
• l est important de rappeler que le « marché » de l’IA générative représente pour le moment un C.A. cumulé encore inférieur à celui du segment des AirPods
• Ceci explique l’approche minimaliste de l’IA générative, le but est de limiter les frais d’exploitation tout en proposant la meilleure expérience possible
• tous les regards sont braqués sur OpenAI, la locomotive de l’IA générative, mais Apple vient d’exposer une stratégie très différente pour monétiser les modèles génératifs.

6 more annotations...

• Slack’s report showed that AI users’ work-life balance was 18% higher, they were 24% more satisfied at work, and 23% better at managing stress, and 29% more likely to say they feel highly passionate about their work.
• The key is “giving people that time and space to learn about this new tool,
• the actual specifics around AI use need to come from individual departments.
• It’s important to give every line of the business their own space to figure out how can I effectively use AI” – “it wouldn’t make sense coming from HR, or the CEO, it has to come from the subject matter experts”,

2 more annotations...

• Our research shows that employees can only tolerate so much toil (defined as work they don’t enjoy) in their day-to-day roles. Beyond four hours a week, people start thinking about leaving their job.
• As part of this study, participants were offered a portfolio of gen AI tools to help with their day-to-day work, as well as access to training materials to help them use these tools. (The technology and training were delivered differently for our control and experiment groups, which we’ll discuss below.) Our early findings suggest practical ways that companies can encourage adoption, enhance enjoyment, and decrease toil
• The true catalyst for driving impactful gen AI adoption lies in the leadership of managers. Managers are essential in navigating change and enabling teams to shift their ways of working to effectively embed new technologies into daily workflows.
• Managers in the top quartile had spent time experimenting with and using the technology 229% more in one month than managers in the bottom quartile.
• As one manager told us, “I champion, cheerlead, and conduct regular pulse checks to understand how their experience with the tool is going and start meetings showing what I personally used the tool for.”
• On average, employees who reported that their managers care about them had 14% higher gen AI usage than employees who do not share this belief.
• The best managers help employees understand the impact and purpose of the activities they do, especially when any change is required. From the start of our experiment, we saw a 66% difference in reported gen AI usage between participants who believed they needed it for their work and those who did not.
• The primary difference between the experimental and control groups was co-creation.
• At the end of the study period, the experimental teams had twice as many active gen AI tool users who spent twice as long using the gen AI tools compared to the control group.
• These results make clear that simply rolling tools out with some light communication and training is not an effective strategy for technology adoption.
• At another organization, our surveys revealed a perfection-oriented culture: a workforce used to academic, peer-reviewed processes. Reliability concerns (reported by 48% of participants) were highlighted as the major reason employees were not embracing gen AI. Employees were afraid that gen AI would make mistakes,

• They report that the biggest obstacle to their effectiveness is the time they spend waiting for others to respond to their availability requests, which creates time pressure resulting in unnecessary toil.

   

  To address this, we introduced new gen AI/AI-driven calendar tools that would allow administrative professionals to easily share availability and to automate these laborious aspects of scheduling.
• In this experiment we were careful to deploy the gen AI tool against the toil (waiting for others to respond) but not the joyful parts (creatively landing the meeting)
• deploy gen AI in ways that deliver productivity without diminishing sources of employee pride and satisfaction.
• Finally, we observed how employees used the time saved with gen AI. Many chose to re-invest it into doing more enjoyable tasks, such as those that involve problem-solving, interpersonal interactions, and learning and development which further enhanced their overall enjoyment at work.

13 more annotations...