The
Daily
Fix
Rest
Garlic Lemon Chicken with Broccolini
Breaking MAD: Generative AI could break the internet
Rest day
Garlic lemon chicken paired with tender-crisp broccolini, seasoned with smoked paprika, oregano, and chili flakes.
A study by Rice University warns that training generative AI models on synthetic data can lead to "Model Autophagy Disorder" (MAD), where AI systems degrade over successive generations.
Enjoy the recovery time, or make-up anything you missed from last week.
Ingredients
8 oz chicken breast, cut into strips
2 Tbsp butter
2 cloves garlic, minced
½ tsp dried oregano
¼ tsp smoked paprika
¼ tsp black pepper
½ lemon, juiced
½ lb broccolini, trimmed (or substitute with regular broccoli florets)
¼ tsp red chili flakes (optional)
½ lemon, sliced
Salt and pepper, to taste
Macronutrients
Protein: 54g
Fat: 30g
Carbs: 16g
Preparation
Marinate the chicken: In a bowl, combine butter (1 Tbsp), minced garlic (2 cloves), oregano (½ tsp), smoked paprika (¼ tsp), black pepper (¼ tsp), salt to taste, and lemon juice (½ lemon). Add the chicken strips (8 oz) and toss to coat. Let the chicken marinate for at least 15-20 minutes.
Cook the chicken: Heat a skillet over medium-high heat. Add the marinated chicken and cook for 3-4 minutes per side, until golden brown and fully cooked through. Remove from the skillet and set aside.
Prepare the broccolini: In the same skillet, add butter (1 Tbsp) if needed. Add the broccolini (½ lb), red chili flakes (¼ tsp, if using), salt, and pepper to taste. Cook for 4-5 minutes, stirring occasionally, until the broccolini is tender-crisp. Add lemon slices (½ lemon) during the last minute of cooking for extra flavor.
Serve: Arrange the cooked chicken on a plate with the sautéed broccolini. Serve with a wedge of lemon for an extra burst of citrus.
This degradation occurs as models trained on AI-generated data develop artifacts, leading to outputs that are progressively marred by errors.
“The problems arise when this synthetic data training is, inevitably, repeated, forming a kind of a feedback loop — what we call an autophagous or 'self-consuming' loop.”
COMMENTS
-----
" developers are already running up against supply limitations and may soon exhaust training resources altogether.
Against this backdrop of data scarcity"
----
We’re not running out of data far from it. Every second, people generate new data across social media, search queries, emails, and every other digital interaction. The problem isn’t that data doesn’t exist; the problem is that it’s expensive to extract, clean, and use in a way that won’t trigger lawsuits. This isn’t a data problem. It’s a money problem
WEDNESDAY 250305