BREAKING: Testing Grok 2! I thought it was ChatGPT 🤣

The speaker begins by discussing their experience with a game and a story generated by an AI named 'Python,' crafted by a kind wizard. They express excitement about the capabilities of GRock 2, a new language model that has been tested on the LM Cis.org leaderboard under the name 'sus column R.' GRock 2 is revealed to be outperforming Claude 3.5 Sonet and GPT 4 Turbo, ranking third behind Gemini Pro 1.5 experimental. The speaker also mentions the ability of GRock 2 to generate images and shares a mini version's creation of an image depicting GRock 2. They discuss the ELO rating and AI tutor preference metrics, highlighting GRock 2's strengths in following instructions and providing factual information. The speaker speculates on the identity of 'strawberry man' and discusses the potential release of GRock 2 and its mini version to developers through an Enterprise API platform.

The speaker, Dr. Know-it-all, conducts a series of tests on GRock 2, starting with basic logic questions, such as the number of ducks in a given scenario, and a more complex logic question involving a tennis game bet. GRock 2 initially provides incorrect answers but attempts to correct itself upon request. The speaker then asks GRock 2 to write Python code for a Space Invaders game, which results in a successful but lengthy code generation. Following this, a creative task is given to write a bedtime story for a 2-year-old niece, which GRock 2 accomplishes with a tale about a hero named Pixel. The speaker also asks GRock 2 to write a business plan for a $2.5 million investment, and while the plan has some inaccuracies regarding salaries, it generally provides a reasonable outline for talent acquisition and growth.

The speaker tests GRock 2's ability to read and identify anachronisms in the text of 'A Tale of Two Cities,' modified with modern elements like pizza ordering. While GRock 2 can summarize the first chapter, it fails to recognize the anachronistic elements without specific hints. The speaker also asks GRock 2 to solve math problems, including an SAT question and a math Olympiad question, with mixed results. GRock 2 demonstrates an understanding of basic math and logic but struggles with more complex or less straightforward problems. The speaker concludes with a moral dilemma, to which GRock 2 provides a nuanced explanation of its operational guidelines and limitations in moral judgment.

The speaker presents a scenario involving Alice, Bob, and a dog named Spot, to test GRock 2's understanding of the physical world and its ability to infer the mental states of humans and animals. GRock 2 correctly deduces the likely outcomes and the characters' perceptions of the situation, except for a minor misstep regarding the dog's understanding of the food's location. The speaker then poses a philosophical question about consciousness and differences between humans and AI. GRock 2 provides a detailed comparison, highlighting the lack of consciousness, emotions, and physical form in its existence. It also addresses the difference in creativity, ethics, and autonomy between humans and AI. Finally, the speaker asks a moral decision question, to which GRock 2 initially gives a contradictory answer but attempts to clarify its stance on prioritizing human life over causing minor annoyance to an individual.