- Activation Function
- Confusion Matrix
- Convolutional Neural Networks
- Forward Propagation
- Generative Adversarial Network
- Gradient Descent
- Linear Regression
- Logistic Regression
- Machine Learning Algorithms
- Multilayer Perceptron
- Naive Bayes
- Neural Networking and Deep Learning
- RuleFit
- Stack Ensemble
- Word2Vec
- XGBoost
- Attention Mechanism
- BERT
- Binary Classification
- Classify Token ([CLS])
- Conversational Response Generation
- GLUE (General Language Understanding Evaluation)
- GPT (Generative Pre-Trained Transformers)
- Language Modeling
- Layer Normalization
- Mask Token ([MASK])
- Probability Distribution
- Probing Classifiers
- SQuAD (Stanford Question Answering Dataset)
- Self-attention
- Separate token ([SEP])
- Sequence-to-sequence Language Generation
- Sequential Text Spans
- Text Classification
- Text Generation
- Transformer Architecture
- WordPiece
- AUC-ROC
- Analytical Review
- Autoencoders
- Bias-Variance Tradeoff
- Decision Optimization
- Explanatory Variables
- Exponential Smoothing
- Level of Granularity
- Long Short-Term Memory
- Loss Function
- Model Management
- Precision and Recall
- Predictive Learning
- ROC Curve
- Recommendation system
- Stochastic Gradient Descent
- Target Leakage
- Target Variable
- Underwriting
A
C
D
G
L
M
N
P
R
S
T
X
- Activation Function
- Confusion Matrix
- Convolutional Neural Networks
- Forward Propagation
- Generative Adversarial Network
- Gradient Descent
- Linear Regression
- Logistic Regression
- Machine Learning Algorithms
- Multilayer Perceptron
- Naive Bayes
- Neural Networking and Deep Learning
- RuleFit
- Stack Ensemble
- Word2Vec
- XGBoost
- Attention Mechanism
- BERT
- Binary Classification
- Classify Token ([CLS])
- Conversational Response Generation
- GLUE (General Language Understanding Evaluation)
- GPT (Generative Pre-Trained Transformers)
- Language Modeling
- Layer Normalization
- Mask Token ([MASK])
- Probability Distribution
- Probing Classifiers
- SQuAD (Stanford Question Answering Dataset)
- Self-attention
- Separate token ([SEP])
- Sequence-to-sequence Language Generation
- Sequential Text Spans
- Text Classification
- Text Generation
- Transformer Architecture
- WordPiece
- AUC-ROC
- Analytical Review
- Autoencoders
- Bias-Variance Tradeoff
- Decision Optimization
- Explanatory Variables
- Exponential Smoothing
- Level of Granularity
- Long Short-Term Memory
- Loss Function
- Model Management
- Precision and Recall
- Predictive Learning
- ROC Curve
- Recommendation system
- Stochastic Gradient Descent
- Target Leakage
- Target Variable
- Underwriting
What is an AI Winter?
There is a natural pattern of societal interest and disinterest when it comes to the development of artificial intelligence (AI). AI winters are the period of time when interest and funding of AI decreases. While interest and funding in AI doesn’t stop altogether during an AI winter, it is severely reduced.
What Causes an AI Winter?
There are many factors that go into the development of an AI winter. Usually there is not one sole factor that causes an AI winter, typically it is a culmination of factors. They are:
Hype
The natural ebb and flow of the hype cycle is a major contributor to AI winters. When organizations promise AI features they cannot deliver, the general public loses interest. If people lose interest, funding dries up, and development stops. Alternatively, if an organization does produce what they promise and it does not live up to the hype, interest will stagnate.
Institutional factors
Oftentimes, the development of AI comes through universities or similar institutions.This significantly impacts AI developers, as they are unable to control funding changes. This is significant because AI developers are not in charge of funding. If budget cuts are made, they must stop development.
Economic factors
The economic status of the country has a major influence on resource allocation to AI development. During war times, recessions, and pandemics, resources must be reallocated to more significant causes.
History of AI Winters
The first AI winter
In the late 1960s, development of perceptrons and connectionism was new and exciting. The majority of developmental funding came from the Defense Advanced Research Projects Agency (DARPA). However, approval of the Mansfield Amendment, which would have required DARPA to fund research directed by a specific mission, caused them to pull their funding of new research. This led to the first notable AI winter in the 1970s.
In 1973, the Parliament of the United Kingdom asked James Lighthill to review the state of AI in the UK. Lighthill criticized the achievement of AI and declared that anything being done in the study of AI could be done in other sciences. Additionally, Lighthill reported that the discoveries have not lived up to the hype surrounding them.
AI was relatively new in the 60s and 70s and hopes were high that it would be a life changing development. However, without funding from DARPA and the release of the Lighthill report, the hype surrounding AI could not sustain the development and the world fell into an AI winter.
The second AI winter
In the 1980s, LISP machines, specialized computers, were built to process the programming language LISP. After three years of development, the market for LISP hardware collapsed. Other organizations continued to sell versions of LISP machines, but ultimately, they all failed.
In 1983, DARPA began funding AI research again. DARPA supported the Strategic Computing Initiative (SCI), which set achievable and realistic goals for development. The SCI was under the direction of the Information Processing Technology Office (IPTO) which had spent over $100 million on 92 different projects in just two years. Eventually, Jack Schwarz, the leader of IPTO stated that AI was not “the next big thing” in the modern world. Once again, the hype surrounding AI dropped, DARPA pulled their funding, and the world fell into a second AI winter.
What Is An AI Summer?
An AI summer is when the development of and interest in AI is booming. There are high expectations set for technological breakthroughs and funding is free flowing. An AI summer is the metaphorical peak of interest. Most, if not all, significant advancements in AI will occur during an AI summer.
The Future of AI
There are many developments in AI that must occur before it is widely accepted for everyday usage. With certain ethical concerns and current AI limitations, an AI winter is possible. However, following the second AI winter, development of AI has been on a sustainable upward trajectory. Recently there have been significant advances in AI including Google Translate, image recognition like Google Image Search, and game-playing systems like AlphaZero and Watson. To view more AI developments, click here.