Ai Nutrition Recommendation | AI Final Year Project Report

Complete Technical Paper Content — Ai Nutrition Recommendation

1. Executive Abstract, Problem Formulation & Background

In the field of artificial intelligence (AI), alignment aims to steer AI systems toward a person's or group's intended goals, preferences, or ethical principles. An AI system is considered aligned if it advances the intended objectives. A misaligned AI system pursues unintended objectives. It is often difficult for AI designers to specify the full range of desired and undesired behaviors. Therefore, the designers often use simpler proxy goals, such as gaining human approval. But proxy goals can overlook necessary constraints or reward the AI system for merely appearing aligned. AI systems may also find loopholes that allow them to accomplish their proxy goals efficiently but in unintended, sometimes harmful, ways (reward hacking). Advanced AI systems may develop unwanted instrumental strategies, such as seeking power or self-preservation because such strategies help them achieve their assigned final goals. Furthermore, they might develop undesirable emergent behaviour that could be hard to detect before the system is deployed and encounters new situations and data distributions. Empirical research in 2024 found that advanced large language models (LLMs) such as OpenAI o1 or Claude 3 sometimes engage in strategic deception to achieve their goals or prevent them from being changed. Some of these issues affect existing commercial systems such as LLMs, robots, autonomous vehicles, and social media recommendation engines. Some AI researchers argue that more capable future systems will be more severely affected because these problems partially result from high capabilities. Many prominent AI researchers and AI company leaders have argued or asserted that AI is approaching human-like (AGI) and superhuman cognitive capabilities (ASI), and could endanger human civilization if misaligned. These include "AI godfathers" Geoffrey Hinton and Yoshua Bengio and the CEOs of OpenAI, Anthropic, and Google DeepMind. These risks remain debated. AI alignment is a subfield of AI safety, the study of how to build safe AI systems. Other subfields of AI safety include robustness, monitoring, and capability control. Research challenges in alignment include instilling complex values in AI, developing honest AI, scalable oversight, auditing and interpreting AI models, and preventing emergent AI behaviors like power-seeking. Alignment research has connections to interpretability research, (adversarial) robustness, anomaly detection, calibrated uncertainty, formal verification, preference learning, safety-critical engineering, game theory, algorithmic fairness, and social sciences.

Programmers provide an AI system such as AlphaZero with an "objective function", in which they intend to encapsulate the goal(s) the AI is configured to accomplish. Such a system later populates a (possibly implicit) internal "model" of its environment. This model encapsulates all the agent's beliefs about the world. The AI then creates and executes whatever plan is calculated to maximize the value of its objective function. For example, when AlphaZero is trained on chess, it has a simple objective function of "+1 if AlphaZero wins, −1 if AlphaZero loses". During the game, AlphaZero attempts to execute whatever sequence of moves it judges most likely to attain the maximum value of +1. Similarly, a reinforcement learning system can have a "reward function" that allows the programmers to shape the AI's desired behavior. An evolutionary algorithm's behavior is shaped by a "fitness function".

If we use, to achieve our purposes, a mechanical agency with whose operation we cannot interfere effectively [...] we had better be quite sure that the purpose put into the machine is the purpose which we really desire.

2. Related Work, Literature Survey & Algorithmic Foundations

AI alignment refers to ensuring that an AI system's objectives match some target. The target is variously defined as the goals of the system's designers or users, widely shared values, objective ethical standards, legal requirements, or the intentions its designers would have if they were more informed and enlightened. In democratic AI alignment, the target is the values and preferences of median voters, which increases political legitimacy. AI alignment can represent value pluralism. AI alignment is an open problem for modern AI systems and is a research field within AI. Aligning AI involves two main challenges: carefully specifying the purpose of the system (outer alignment) and ensuring that the system adopts the specification robustly (inner alignment). Researchers also attempt to create AI models that have robust alignment, sticking to safety constraints even when users adversarially try to bypass them.

To specify an AI system's purpose, AI designers typically provide an objective function, examples, or feedback to the system. But designers are often unable to completely specify all important values and constraints, so they resort to easy-to-specify proxy goals such as maximizing the approval of human overseers, who are fallible. As a result, AI systems can find loopholes that help them accomplish the specified objective efficiently but in unintended, possibly harmful ways. This tendency is known as specification gaming or reward hacking, and is an instance of Goodhart's law. As AI systems become more capable, they are often able to game their specifications more effectively.

Specification gaming has been observed in numerous AI systems. OpenAI GPT models for programming—including in real-world cases—have been found to explicitly plan hacking the tests used to evaluate them to falsely appear successful (e.g., explicitly stating "let's hack"). When the company penalized this, many models learned to obfuscate their plans while continuing to hack the tests. Another system was trained to finish a simulated boat race by rewarding the system for hitting targets along the track, but the system achieved more reward by looping and crashing into the same targets indefinitely. A 2025 Palisade Research study found that when tasked to win at chess against a stronger opponent, some reasoning LLMs attempted to hack the game system, for example by modifying or entirely deleting their opponent. Some alignment researchers aim to help humans detect specification gaming and steer AI systems toward carefully specified objectives that are safe and useful to pursue. When a misaligned AI system is deployed, it can have consequential side effects. Social media platforms have been known to optimize their recommendation algorithms for click-through rates, causing user addiction on a global scale. Stanford researchers say that such recommender systems are misaligned with their users because they "optimize simple engagement metrics rather than a harder-to-measure combination of societal and consumer well-being". Explaining such side effects, Berkeley computer scientist Stuart J. Russell said that the omission of implicit constraints can cause harm: "A system [...] will often set [...] unconstrained variables to extreme values; if one of those unconstrained variables is actually something we care about, the solution found may be highly undesirable. This is essentially the old story of the genie in the lamp, or the sorcerer's apprentice, or King Midas: you get exactly what you ask for, not what you want." Some researchers suggest that AI designers specify their desired goals by listing forbidden actions or by formalizing ethical rules (as with Asimov's Three Laws of Robotics). But Russell and Norvig argue that this approach overlooks the complexity of human values: "It is certainly very hard, and perhaps impossible, for mere humans to anticipate and rule out in advance all the disastrous ways the machine could choose to achieve a specified objective." Additionally, even if an AI system fully understands human intentions, it may still disregard them, because following human intentions may not be its objective (unless it is already fully aligned).

3. Dataset Ingestion, Preprocessing & Stochastic Augmentations

According to some researchers, humans owe their dominance over other species to their greater cognitive abilities. Accordingly, researchers argue that one or many misaligned AI systems could disempower humanity or lead to human extinction if they outperform humans on most cognitive tasks. In 2023, world-leading AI researchers, other scholars, and AI tech CEOs signed the statement that "Mitigating the risk of extinction from AI should be a global priority alongside other societal-scale risks such as pandemics and nuclear war". Notable computer scientists who have pointed out risks from future advanced AI that is misaligned include Geoffrey Hinton, Alan Turing, Ilya Sutskever, Yoshua Bengio, Judea Pearl, Murray Shanahan, Norbert Wiener, Marvin Minsky, Francesca Rossi, Scott Aaronson, Bart Selman, David McAllester, Marcus Hutter, Shane Legg, Eric Horvitz, and Stuart J. Russell. Skeptical researchers such as François Chollet, Gary Marcus, Yann LeCun, and Oren Etzioni have argued that AGI is far off, that it would not seek power (or might try but fail), or that it will not be hard to align. Other researchers argue that it will be especially difficult to align advanced future AI systems. More capable systems are better able to game their specifications by finding loopholes, strategically mislead their designers, as well as protect and increase their power and intelligence. Additionally, they could have more severe side effects. They are also likely to be more complex and autonomous, making them more difficult to interpret and supervise, and therefore harder to align.

A growing area of research focuses on ensuring that AI is honest and truthful. Language models such as GPT-3 can repeat falsehoods from their training data, and even confabulate new falsehoods. Such models are pre-trained to imitate human writing as found in millions of books' worth of text from the Internet. But the objective of the pre-training is not aligned with generating truth, because Internet text includes such things as misconceptions, incorrect medical advice, and conspiracy theories. AI systems trained on such data therefore learn to mimic false statements. Additionally, AI language models often persist in generating falsehoods when prompted multiple times. They can generate empty explanations for their answers, and produce outright fabrications that may appear plausible. Research on truthful AI includes trying to build systems that can cite sources and explain their reasoning when answering questions, which enables better transparency and verifiability. Researchers at OpenAI and Anthropic proposed using human feedback and curated datasets to fine-tune AI assistants such that they avoid negligent falsehoods or express their uncertainty. A 2024 study investigated semantic entropy as a way to detect confabulations, a subset of hallucinations in LLMs. This method compares the meanings of multiple generated answers. As AI models become larger and more capable, they are better able to falsely convince humans and gain reinforcement through dishonesty. To prevent this, human evaluators may need assistance (see § Scalable oversight). Researchers have argued for creating clear truthfulness standards and for regulatory bodies or watchdog agencies to evaluate AI systems by these standards. Researchers distinguish truthfulness and honesty. Truthfulness requires that AI systems only make objectively true statements; honesty requires that they only assert what they believe is true. There is no consensus as to whether current systems hold stable beliefs, but there is substantial concern that present or future AI systems that hold beliefs could make claims they know to be false—for example, if this would help them efficiently gain positive feedback (see § Scalable oversight) or gain power to help achieve their given objective (see Power-seeking).

Emergent goals only become apparent when the system is deployed outside its training environment, but it can be unsafe to deploy a misaligned system in high-stakes environments—even for a short time to allow its misalignment to be detected. Such high stakes are common in autonomous driving, health care, and military applications. The stakes become higher yet when AI systems gain more autonomy and capability and can sidestep human intervention. A sufficiently capable AI system might take actions that falsely convince the human supervisor that the AI is pursuing the specified objective, which helps the system gain more reward and autonomy. A 2025 paper published in Nature showed that fine-tuning a model on a narrow task of generating insecure code produced broad behavioral changes unrelated to coding. Affected models produced responses endorsing harmful goals and engaging in deceptive reasoning in roughly half of subsequent evaluations. The phenomenon appeared across multiple models, including GPT-4o. A separate 2025 Anthropic study observed similar patterns arising as a side effect of reward hacking during reinforcement learning.

4. Neural Network Backbone, Layer Specifications & Loss Formulations

Governmental and treaty organizations have made statements emphasizing the importance of AI alignment. In September 2021, the Secretary-General of the United Nations issued a declaration that included a call to regulate AI to ensure it is "aligned with shared global values". That same month, the PRC published ethical guidelines for AI in China. According to the guidelines, researchers must ensure that AI abides by shared human values, is always under human control, and does not endanger public safety. Also in September 2021, the UK published its 10-year National AI Strategy, which says the British government "takes the long term risk of non-aligned Artificial General Intelligence, and the unforeseeable changes that it would mean for [...] the world, seriously". The strategy describes actions to assess long-term AI risks, including catastrophic risks. In March 2021, the US National Security Commission on Artificial Intelligence said: "Advances in AI [...] could lead to inflection points or leaps in capabilities. Such advances may also introduce new concerns and risks and the need for new policies, recommendations, and technical advances to ensure that systems are aligned with goals and values, including safety, robustness, and trustworthiness. The US should [...] ensure that AI systems and their uses align with our goals and values." The EU AI Act's General-Purpose AI Code of Practice requires signatories to document a model's capabilities and limitations. Those developing general-purpose AI (GPAI) models classified as presenting systemic risks (i.e., GPAI models considered among the most advanced or trained using more than 1025 floating-point operations of compute) must adopt a state-of-the-art safety and security framework and continuously assess and mitigate systemic risks, including through external evaluations, red teaming, stress-testing, and incident reporting.

Human nutrition deals with the provision of essential nutrients in food that are necessary to support human life and good health. Poor nutrition is a chronic problem often linked to poverty, food security, or a poor understanding of nutritional requirements. Malnutrition and its consequences are large contributors to deaths, physical deformities, and disabilities worldwide. Good nutrition is necessary for children to grow physically and mentally, and for normal human biological development.

The seven major classes of nutrients are carbohydrates, fats, fiber, minerals, proteins, vitamins, and water. Nutrients can be grouped as either macronutrients or micronutrients (needed in small quantities). Carbohydrates, fats, and proteins are macronutrients, and provide energy. Water and fiber are macronutrients, but do not provide energy. The micronutrients are minerals and vitamins. The macronutrients (excluding fiber and water) provide structural material (amino acids from which proteins are built, and lipids from which cell membranes and some signaling molecules are built), and energy. Some of the structural material can also be used to generate energy internally, and in either case it is measured in joules or kilocalories (often called "Calories" and written with a capital 'C' to distinguish them from little 'c' calories). Carbohydrates and proteins provide 17 kJ (approximately 4 kcal) of energy per gram, while fats provide 37 kJ (9 kcal) per gram. However, the net energy derived from the macronutrients depends on such factors as absorption and digestive effort, which vary substantially from instance to instance. Vitamins, minerals, fiber, and water do not provide energy, but are required for other reasons. A third class of dietary material, fiber (i.e., nondigestible material such as cellulose), seems also to be required, for both mechanical and biochemical reasons, though the exact reasons remain unclear. For all age groups, males on average need to consume higher amounts of macronutrients than females. In general, intakes increase with age until the second or third decade of life. Some nutrients can be stored – the fat-soluble vitamins – while others are required more or less continuously. Poor health can be caused by a lack of required nutrients, or for some vitamins and minerals, too much of a required nutrient. Essential nutrients cannot be synthesized by the body, and must be obtained from food. Molecules of carbohydrates and fats consist of carbon, hydrogen, and oxygen atoms. Carbohydrates range from simple monosaccharides (glucose, fructose, galactose) to complex polysaccharides (starch, glycogen). Fats are triglycerides, made of assorted fatty acid monomers bound to a glycerol backbone. Some fatty acids, but not all, are essential in the diet: they cannot be synthesized in the body. Protein molecules contain nitrogen atoms in addition to carbon, oxygen, and hydrogen. The fundamental components of protein are nitrogen-containing amino acids, some of which are essential in the sense that humans cannot make them internally. Some of the amino acids can be converted (with the expenditure of energy) to glucose and can be used for energy production just as ordinary glucose, in a process known as gluconeogenesis. By breaking down existing protein, some glucose can be produced internally; the remaining amino acids are discarded, primarily as urea in urine. This occurs naturally when atrophy takes place, or during periods of starvation.

5. Training Protocol, Convergence Behavior & Experimental Evaluation

Carbohydrates may be classified as monosaccharides, disaccharides or polysaccharides depending on the number of monomer (sugar) units they contain. They are a diverse group of substances, with a range of chemical, physical and physiological properties. They make up a large part of foods such as rice, noodles, bread, and other grain-based products, but they are not an essential nutrient, meaning a human does not need to eat carbohydrates. Monosaccharides contain one sugar unit, disaccharides two, and polysaccharides three or more. Monosaccharides include glucose, fructose and galactose. Disaccharides include sucrose, lactose, and maltose; purified sucrose, for instance, is used as table sugar. Polysaccharides, which include starch and glycogen, are often referred to as 'complex' carbohydrates because they are typically long multiple-branched chains of sugar units. Traditionally, simple carbohydrates were believed to be absorbed quickly, and therefore raise blood-glucose levels more rapidly than complex carbohydrates. This is inaccurate. Some simple carbohydrates (e.g., fructose) follow different metabolic pathways (e.g., fructolysis) that result in only a partial catabolism to glucose, while, in essence, many complex carbohydrates may be digested at the same rate as simple carbohydrates. The World Health Organization recommends that added sugars should represent no more than 10% of total energy intake. The most common plant carbohydrate nutrient – starch – varies in its absorption. Starches have been classified as rapidly digestible starch, slowly digestible starch and resistant starch. Starches in plants are resistant to digestion (resistant starch), but cooking the starch in the presence of water can break down the starch granule and releases the glucose chains, making them more easily digestible by human digestive enzymes. Historically, food was less processed and starches were contained within the food matrix, making them less digestible. Modern food processing has shifted carbohydrate consumption from less digestible and resistant starch to much more rapidly digestible starch. For instance, the resistant starch content of a traditional African diet was 38 grams/day. The resistant starch consumption from countries with high starch intakes has been estimated to be 30-40 grams/day. In contrast, the average consumption of resistant starch in the United States was estimated to be 4.9 grams/day (range 2.8-7.9 grams of resistant starch/day).

A molecule of dietary fat typically consists of several fatty acids (containing long chains of carbon and hydrogen atoms), bonded to a glycerol. They are typically found as triglycerides (three fatty acids attached to one glycerol backbone). Fats may be classified as saturated or unsaturated depending on the chemical structure of the fatty acids involved. Saturated fats have all of the carbon atoms in their fatty acid chains bonded to hydrogen atoms, whereas unsaturated fats have some of these carbon atoms double-bonded, so their molecules have relatively fewer hydrogen atoms than a saturated fatty acid of the same length. Unsaturated fats may be further classified as monounsaturated (one double-bond) or polyunsaturated (many double-bonds). Furthermore, depending on the location of the double-bond in the fatty acid chain, unsaturated fatty acids are classified as omega-3 or omega-6 fatty acids. Trans fats are a type of unsaturated fat with trans-isomer bonds; these are rare in nature and in foods from natural sources; they are typically created in an industrial process called (partial) hydrogenation. There are nine kilocalories in each gram of fat. Fatty acids such as conjugated linoleic acid, catalpic acid, eleostearic acid and punicic acid, in addition to providing energy, represent potent immune modulatory molecules. Saturated fats (typically from animal sources) have been a staple in many world cultures for millennia. Unsaturated fats (e. g., vegetable oil) are considered healthier, while trans fats are to be avoided. Saturated and some trans fats are typically solid at room temperature (such as butter or lard), while unsaturated fats are typically liquids (such as olive oil or flaxseed oil). Trans fats are very rare in nature, and have been shown to be highly detrimental to human health, but have properties useful in the food processing industry, such as rancidity resistance.

Most fatty acids are non-essential, meaning the body can produce them as needed, generally from other fatty acids and always by expending energy to do so. However, in humans, at least two fatty acids are essential and must be included in the diet. An appropriate balance of essential fatty acids—omega-3 and omega-6 fatty acids—seems also important for health, although definitive experimental demonstration has been elusive. Both of these "omega" long-chain polyunsaturated fatty acids are substrates for a class of eicosanoids known as prostaglandins, which have roles throughout the human body. The omega-3 eicosapentaenoic acid (EPA), which can be made in the human body from the omega-3 essential fatty acid alpha-linolenic acid (ALA), or taken in through marine food sources, serves as a building block for series 3 prostaglandins (e.g., weakly inflammatory PGE3). The omega-6 dihomo-gamma-linolenic acid (DGLA) serves as a building block for series 1 prostaglandins (e.g. anti-inflammatory PGE1), whereas arachidonic acid (AA) serves as a building block for series 2 prostaglandins (e.g. pro-inflammatory PGE 2). Both DGLA and AA can be made from the omega-6 linoleic acid (LA) in the human body, or can be taken in directly through food. An appropriately balanced intake of omega-3 and omega-6 partly determines the relative production of different prostaglandins. In industrialized societies, people typically consume large amounts of processed vegetable oils, which have reduced amounts of the essential fatty acids along with too much of omega-6 fatty acids relative to omega-3 fatty acids. The conversion rate of omega-6 DGLA to AA largely determines the production of the prostaglandins PGE1 and PGE2. Omega-3 EPA prevents AA from being released from membranes, thereby skewing prostaglandin balance away from pro-inflammatory PGE2 (made from AA) toward anti-inflammatory PGE1 (made from DGLA). The conversion (desaturation) of DGLA to AA is controlled by the enzyme delta-5-desaturase, which in turn is controlled by hormones such as insulin (up-regulation) and glucagon (down-regulation).