| # | Unlocked | Locked | Name | Beschreibung | Spieler | Global |
|---|---|---|---|---|---|---|
| 1 |
|
| Learning Plateau | MCG falls to bottom, no agents ascend | 1 | 30.5% |
| 2 |
|
| NaN Trap | All Agents fall off the planet | 1 | 28.9% |
| 3 |
|
| Adam Optimizer | Have a Reinforcement Learning Agent survive to the end of the film | 0 | 26% |
| 4 |
|
| Loss Minimization | Grab an Agent while they’re falling during the fall moment | 0 | 23.6% |
| 5 |
|
| Activation Function | Achieve the Good Ending | 0 | 19.6% |
| 6 |
|
| Normalization | Have all 5 agents consume in a single world. | 0 | 17.8% |
| 7 |
|
| Unsupervised Learning | Achieve a good ending without planting a MacGuffin | 0 | 17.3% |
| 8 |
|
| Gradient Descent | Descend deep into the simulation | 0 | 17.3% |
| 9 |
|
| A.G.I. | All agents ascend | 0 | 17.2% |
| 10 |
|
| Turing Test | Have a Reinforcement Learning Agent to ascend to the next planet | 0 | 16.9% |
| 11 |
|
| Computer Vision | See all moments | 0 | 16.7% |
| 12 |
|
| Decision Boundary | Pick up 20 Agents that are very aggressive | 0 | 16.5% |
| 13 |
|
| Backpropagation | Achieve a good ending 3 times in a row without planting any MacGuffins | 0 | 16.2% |
| 14 |
|
| Weights and Biases | Cover the entire surface area of the planet with MacGuffin roots | 0 | 16.2% |
| 15 |
|
| One-hot Encoding | One interaction good ending | 0 | 16.2% |
| 16 |
|
| Singularity | 5 RL agents ascend | 0 | 16.2% |