In this first installment of PAI’s Summer School Series, we’re breaking down the basics of artificial intelligence. Whether you are an AI pro or just beginning to explore it, this primer will help sharpen your understanding of the technology changing the world.
What is AI?
Human intelligence is defined as the ability to learn, reason, and apply knowledge or skills to solve problems. While Artificial Intelligence, or AI, is not like human intelligence, it is designed to simulate it. There’s a very important distinction between humans and AI, and that is that AI systems do not “think,” but rather reason and solve problems based on how they are trained. AI systems are built to process information, recognize patterns, and make decisions on a much larger scale and at a faster rate than humans.
There are different types of AI. Narrow AI, or “weak” AI, is the only kind of AI that exists today. It is designed to handle and manage very specific tasks and functions like recommending certain kinds of videos on social media or generating text (via AI assistants like ChatGPT). It can do many specific tasks very well but it can’t generalize or think beyond its programming. You may hear the terms “AGI” and “Superintelligence” being used as people speculate about the future capabilities of AI, but as of now, they are both hypothetical types of AI.
Common AI Terms You Might Have Heard
Many people first heard of AI in 2022, when Open AI first released their generative AI chatbot “ChatGPT”. While not a new concept, this AI tool became a viral sensation due to its accessibility and its impressive ability to “understand” and generate human-like text. Since the initial release of GPT-3.5, AI technology has taken center stage in the tech field, flooding the media with jargon and confusing terms. So what do these common AI terms even mean?
Foundation Model: Foundation models are AI systems with generally applicable functions that are designed to be used across a variety of contexts. The current generation of these systems is characterized by training deep learning models on large datasets (which requires significant computational resources) to perform numerous tasks that can serve as the “foundation” for a wide array of downstream applications.
Large Language Model (LLM): LLMs are a type of foundation model, primarily used to program systems, like generative AI systems. They are trained on very large sets of data and use machine learning techniques to improve and refine themselves.
Generative AI: Generative AI is a kind of AI system that can produce text, images, audio, and video. Examples of generative AI systems are Claude, Gemini, and ChatGPT, Midjourney, and Sora. These AI systems work by generating Synthetic Media when a user prompts the system with a specific request.
Agentic AI: This kind of AI system can act on behalf of users, with some degree of autonomy, to achieve goals without human intervention or guidance. The goals of agents are to understand a user’s general goals and utilize context to solve specific problems without explicit instructions. For example, where ChatGPT can generate a pizza recipe for you, if asked, an AI agent can find the best pizza place near you, look for and book a reservation, and schedule the reservation for when you are available in the week.
Although generative AI has become the most popular form of AI, there’s more to it. AI has actually been around since 1956, but the concepts which led to the development of AI have been around for much longer, with some theorizing it all started as far back as the eighteenth century. AI is in many of the applications and systems you interact with on a daily basis. When you apply for a car loan, an AI system is running in the background to weigh whether or not you should qualify. When you are scrolling through Netflix to pick your next binge-watch, an AI system is running in the background to decide which shows or movies to recommend to you based on your watch history. When you go on a roadtrip and use Google or Apple maps to help you navigate, an AI system is running in the background to optimize your route and avoid traffic. AI is everywhere, but how do these machines know what they know?
Machine learning: Machine learning refers to the method in which AI “learns” over time. Incorporating computer science, math, and coding, this process involves the development of algorithms that help machines learn without any human assistance.
Algorithm: Algorithms are instructions that tell the computer how to make decisions, perform tasks, or execute a function autonomously. Algorithms look for patterns in data and over time, as they work, it improves itself.
While much of this work sounds like computers are teaching computers how to “think” and “operate,” humans play a very critical role in the development of AI systems. Apart from overseeing the development of these systems, there are hundreds of millions of people around the world who collect the data that powers AI and train these machines to identify and recognize patterns in the data. There are also millions of people around the world who are dedicated to understanding the impacts these systems have on people and society.
Data Workers:Data workers are individuals who perform data enrichment tasks, such as cleaning, labeling, and moderating large datasets, that are crucial for training machine learning models, especially those powering AI systems.
Bias: Bias in AI refers to a systematic error that leads to unfair outcomes. Bias is typically introduced through human error in programming, data collection, or training. Bias in AI systems can exacerbate preexisting risks posed to marginalized groups.
Responsible AI
While AI has become a powerful tool in our everyday lives it is also important to recognize that the ubiquity of it means it can also have great potential for harm or misuse. Used irresponsibly, AI can amplify and reinforce discrimination, violate privacy, and even be used to spread false information. That is why Partnership on AI is dedicated to advancing the responsible development, deployment, and use of AI systems. AI is a tool that can be leveraged for a multitude of purposes, but it should always benefit society. To learn more about how we advance responsible AI, sign up for our newsletter.
The world today is war-torn, starting with Russia’s attacks on Ukraine to Israel’s devastation in Palestine and now in Iran, putting the entire West Asia in jeopardy.
The geometrics of war has completely changed, from Blitzkrieg (lightning war) in World War II to the use of sophisticated and technologically driven missiles in these latest armed conflicts. The most recent wars are being driven by use of artificial intelligence (AI) to narrow down potential targets.
There have been multiple evidences which indicate that Israeli forces have deployed novel AI-driven targeting tools in Gaza. One system, nicknamed “Lavender” is an AI-enabled database that assigns risk scores to Gazans based on patterns in their personal data (communication, social connections) to identify “suspected Hamas or Islamic Jihad operatives”. Lavender has flagged up to 37,000 Palestinians as potential targets early in the war.
A second system, “Where is Daddy?”, uses mobile phone location tracking to notify operators when a marked individual is at home. The initial strikes using these automated generated systems targeted individuals in their private homes on the pretext of targeting the terrorists. But innocent women and young children also lost their lives in these attacks. This technology was developed as a replacement of human acumen and strategy to identify and target the suspects.
According to the Humans Rights Watch report (2024), around 70 per cent of people who have lost lives were women and children. The United Nations agency has also verified the details of 8,119 victims killed in Gaza from November 2023 to April 2024. The report showed that 44 per cent of the victims were children and 26 per cent were women. The humans are merely at the mercy of this sophisticated technology that identified the suspected militants and targeted them.
The use of AI-based tools like “Lavender” and “Where’s Daddy?” by Israel in its war against Palestine raises serious questions about the commitment of countries to the international legal framework and the ethics of war. Use of such sophisticated AI targeted tools puts the weaker nations at the dictate of the powerful nations who can use these technologies to inflict suffering for the non-combatants.
The international humanitarian law (IHL) and international human rights law (IHRL) play a critical yet complex role in the context of AI during conflict situations such as the Israel-Palestine Conflict. Such AI-based warfare violates the international legal framework principles of distinction, proportionality and precaution.
The AI systems do not inherently know who is a combatant. Investigations report that Lavender had an error rate on the order of 10 per cent and routinely flagged non-combatants (police, aid workers, people who merely shared a name with militants). The reported practice of pre-authorising dozens of civilian deaths per strike grossly violates the proportionality rule.
An attack is illegal if incidental civilian loss is “excessive” in relation to military gain. For example, one source noted that each kill-list target came with an allowed “collateral damage degree” (often 15–20) regardless of the specific context. Allowing such broad civilian loss per target contradicts IHL’s core balancing test (ICRC Rule 14).
The AI-driven process has eliminated normal safeguards (verification, warnings, retargeting). IHRL continues to apply alongside IHL in armed conflict contexts. In particular, the right to life (ICCPR Article 6) obliges states to prevent arbitrary killing.
The International Court of Justice has held that while the right to life remains in force during war, an “arbitrary deprivation of life” must be assessed by reference to the laws of war. In practice, this means that IHL’s rules become the benchmark for whether killings are lawful.
However, even accepting lex specialis (law overriding general law), the reported AI strikes raise grave human rights concerns especially the Right to Life (ICCPR Art. 6) and Right to Privacy (ICCPR Art. 17).
Ethics of war, called ‘jus in bello’ in the legal parlance, based on the principles of proportionality (anticipated moral cost of war) and differentiation (between combatants and non-combatants) has also been violated. Article 51(5) of Additional Protocol I of the 1977 Geneva Convention said that “an attack is disproportionate, and thus indiscriminate, if it may be expected to cause incidental loss of civilian life, injury to civilians, damage to civilian objects, or a combination thereof, which would be excessive in relation to the concrete and military advantage”.
The Israel Defense Forces have been indiscriminately using AI to target potential targets. These targets though aimed at targeting militants have been extended to the non-military targets also, thus causing casualties to the civilians and non-combatants. Methods used in a war is like a trigger which once warded off is extremely difficult to retract and reconcile. Such unethical action creates more fault lines and any alternate attempt at peace resolution and mediation becomes extremely difficult.
The documented features of systems like Lavender and Where’s Daddy, based on automated kill lists, minimal human oversight, fixed civilian casualty “quotas” and use of imprecise munitions against suspects in homes — appear to contravene the legal and ethical principles.
Unless rigorously constrained, such tools risk turning warfare into arbitrary slaughter of civilians, undermining the core humanitarian goals of IHL and ethics of war. Therefore, it is extremely important to streamline the unregulated use of AI in perpetuating war crimes as it undermines the legal and ethical considerations of humanity at large.
The AI landscape is evolving rapidly, and with the rise of agentic AI, trust has never been more critical. As businesses continue to integrate AI into their operations and customer experiences, leaders must ensure that these technologies are developed and deployed in a responsible manner. Leading with trust and responsibility is not optional. Enterprise customers require this as part of their AI adoption journey, and trust is essential to a future in which AI creates opportunities for everyone.
Salesforce is proud to be one of the first companies to contribute to the reporting framework developed by the OECD under the G7 Hiroshima AI Process (HAIP). Voluntary frameworks like this empower organisations to prioritise ethical practices, transparency, and governance at every stage of AI development and deployment, fostering more trustworthy AI ecosystems and enhancing global alignment on best practices.
Risk identification: Laying the foundation for trustworthy AI
An effective, responsible AI approach begins with a comprehensive strategy for risk identification and evaluation. Organisations should define and classify different types of AI-related risks, particularly those that could cause serious harm. This is especially important in enterprise settings, where AI systems are often tailored and used in various contexts. At Salesforce, the Responsible AI and Tech (RAIT) product managers within our Office of Ethical and Humane Use (OEHU) are central to this effort. During these reviews, RAIT product managers work closely with product teams to understand use cases, technology stacks, and intended audiences. The process involves identifying and categorising potential risks into subtypes of sociotechnical harm, as well as assessing both inherent and residual risks to provide a holistic view of potential impacts, enabling informed decision-making and effective mitigation strategies.
Our AI Acceptable Use Policy provides clear guidance on the uses for which our customers are prohibited from using AI tools. This includes automated decision-making with legal consequences, predictions of an individual’s protected characteristics, or high-risk scenarios that could result in serious harm or injury.
Ongoing risk management: Protecting AI systems in real-time
Responsible AI experts must collaborate closely with product teams at all stages of the innovation process to devise effective mitigation strategies. Standardised guardrails, such as Salesforce’s “trust patterns”, can include features like mindful friction, which introduces checkpoints for thoughtful decision-making, or transparency notifications that inform users when they are interacting with AI systems.
Organisations should also establish comprehensive frameworks that protect data privacy and security throughout every stage of the product development process. Salesforce’s Trust Layer includes functionalities such as secure data handling, zero data retention, ethics by design, an audit trail, and real-time toxicity detection.
Finally, Salesforce has clear evidence from enterprise customers that testing products against trust and safety metrics, such as bias, privacy, and truthfulness, is an important business strategy and benefit. At Salesforce, we regularly introduce red teaming exercises, which simulate potential risks in controlled environments, to identify vulnerabilities and risks within products. Tactics like this are particularly important as autonomous agents become increasingly widespread.
Transparency reporting: Building trust through honest communication and knowledge-sharing
Transparency and honesty are core tenets of our trusted AI principles, which we augmented with our guidelines for trusted generative AI, and remain applicable to the agentic AI era. Organisations should ensure that users and stakeholders are informed about how and when AI is used. At Salesforce, we regularly share information about our product capabilities through our newsroom, blogs, and Trailhead, our free online learning platform.
Salesforce also regularly reports on our progress in responsible AI efforts. Most recently, our Trusted AI and Agents Report explained our approach to designing and deploying AI agents.
Furthermore, we aim to be transparent about the use of personal data. Salesforce enables customers to control how their data is used for AI. Whether using our own Salesforce-hosted models or external models within our shared trust boundary, no context is stored. The large language model forgets both the prompt and the output immediately after processing.
Organisational governance: Embedding responsible AI practices across the company
Gaining the buy-in from all parts of the organisation to deliver a truly effective responsible AI approach is critical. Salesforce embeds AI risk management within its organisational governance framework through various structures and practices. The company’s trusted AI principles, first developed in 2018 and augmented for generative AI in 2023, guide responsible development and deployment, focusing on intentional design and system-level controls.
Our governance infrastructure includes:
The Office of Ethical and Humane Use (OEHU), which regularly interacts with the executive leadership team for policy and product review and approval. The OEHU also leads the Trusted AI Review process to identify, mitigate, and track potential risks early in development.
The AI Trust Council, comprising executives across various departments, aligns and speeds up decision-making for AI products.
The Ethical Use Advisory Council, established in 2018 with external experts and internal executives, provides strategic guidance on product and policy recommendations.
The Cybersecurity and Privacy Committee of the Board of Directors, which meets quarterly with the Chief Ethical and Humane Use Officer to review AI priorities.
The Human Rights Steering Committee, meeting quarterly, oversees the human rights program, including identifying and mitigating salient risks.
A shared commitment to responsible AI: Aligning with global standards
The future of responsible AI depends on a collective commitment to developing systems that are innovative, trustworthy, ethical, and secure. Emphasising transparency and robust governance will unlock AI’s full potential while ensuring the safety of customers and stakeholders.
The G7 HAIP reporting framework provides an effective global benchmark for responsible AI initiatives, providing a structured approach for organisations to manage the risks and benefits of AI technologies. As these frameworks gain widespread adoption, they will promote consistency in responsible AI practices, building greater trust among users and society. Salesforce is committed to working with all stakeholders and navigating this transformative AI era with trust, responsibility, and ethics guiding the way.
“We need to ask ourselves: is AI actually worth the costs?” – SWZ
Immer informiert bleiben
Den SWZ-Newsletter abonnieren
Bozen – Brent Mittelstadt is a professor of data ethics and policy at the Oxford Internet Institute (OII) at the University of Oxford. As a philosopher of technology, he specializes in artificial intelligence ethics. His work critically examines algorithmic decision-making, fairness, transparency, and accountability, influencing AI governance and policy. By integrating multidisciplinary perspecti…
Jetzt Abo abschließen oder einloggen, um den ganzen Artikel zu lesen.
Flexi-Abo
Online | Unbegrenzter Zugang zu allen Inhalten auf www.swz.it
App | Unbegrenzter Zugang in der Mobile und Tablet App (iOS und Android)
ePaper | Die digitale 1:1-Ausgabe der Printversion (schon am Donnerstag ab 20 Uhr)
Diese Website verwendet Cookies während Sie durch die Website navigieren, um Ihre Erfahrungen zu verbessern. Für die Funktionalität der Website verwenden wir technische Cookies und für anonyme Statistiken Marketing Cookies. Durch das Klicken auf Akzeptieren, stimmen sie automatisch den Marketing Cookies zu, außer sie haben diese deaktiviert.
