Bret Victor: The Future of Programming in 1973

Joey Reid

Summary:

Bret Victor, dressed as an IBM systems engineer from 1973, discusses the future of programming at the DBX Conference. [00:00] [Image 1] [Image 2]
He highlights the rapid advancements in computing technology, such as Moore's Law [01:54] [Image 10], contrasting it with the slow pace of adopting new programming ideas.
Victor illustrates this resistance with historical examples: the initial rejection of assemblers and high-level languages like Fortran by programmers accustomed to binary and assembly code [02:34] [Image 14] [Image 18].
He then outlines four key ideas for the future of programming, drawing from "recent research" (presented as innovations from the 1960s-early 1970s) [05:25] [Image 22].
The first idea is direct manipulation of data, exemplified by Ivan Sutherland's Sketchpad from 1962, allowing users to draw and apply constraints visually [06:47] [Image 26] [Image 28].
The second concept is programming with goals and constraints, where the system figures out how to achieve a desired outcome, as seen in Carl Hewitt's Planner [1969] [10:19] [Image 34] and Prolog [1972]. He also highlights the challenge of communicating between independently written programs, using a diagrammatic representation [14:48] [Image 47].
The third idea focuses on spatial representations of information, moving beyond linear text, inspired by Doug Engelbart's NLS [1968] [17:21] [Image 54], RAND Corporation's GRAIL [1968] [18:20] [Image 60], and Larry Tesler's Smalltalk Browser [1972] [19:35] [Image 62]. He emphasizes the importance of immediate response in these interactive systems [20:41] [Image 66].
The fourth concept is concurrent programming, necessary for massively parallel processors [23:33] [Image 76], replacing the sequential Von Neumann architecture [21:57] [Image 73] and the limitations of threads and locks [25:38] [Image 79]. He proposes Carl Hewitt's Actor Model [1973] as a more suitable approach [26:19] [Image 85].
Victor concludes by stating that the most dangerous thought for a creative person is believing they already know everything, encouraging an open mind to new ideas [31:23] [Image 97].

Bret Victor's 1973 Keynote on the Future of Programming [00:00]

Introduction to the DBX Conference 1973 [00:00]
- Bret Victor, dressed as an IBM systems engineer, welcomes the audience of "programmers of automatic computing machines." [Image 2]
- The title slide for the presentation "THE FUTURE OF PROGRAMMING" is displayed. [Image 1] [Image 6]
- He notes the increasing importance of computing in society, with "thousands of computers" now in use across various fields.
- Computers have shrunk significantly from room-sized machines to smaller proportions, reflecting a period of rapid change.
- The talk's premise is to look ahead 40 years to predict the future of programming, given the current knowledge in 1973.
The Nature of Adopting Ideas: Technology vs. Mindset [01:26]
- Technology changes quickly, but people's minds change slowly.
- New technologies are easily adopted, but new ways of thinking face significant resistance.
- Moore's Law [1965] [01:54]
  - Gordon Moore of Intel observed computing capacity increasing exponentially, a trend expected to continue. [Image 10]
  - This technological advancement is often taken for granted.
- Resistance to New Programming Paradigms [02:34]
  - IBM 650 [1953] and Binary Programming [02:34]
    - Early programmers coded in absolute binary, writing numeric codes for each instruction.
  - SOAP (Symbolic Optimizing Assembly Program) [1956] [03:00]
    - Stan Poley invented the assembler, allowing symbolic words and variable names, a more powerful and productive approach. [Image 14]
    - Binary programmers resisted this, not seeing its value, illustrating "people's minds change slowly."
  - John von Neumann's Resistance to Assemblers [03:51]
    - Von Neumann, a renowned scientist, famously saw no need for anything beyond machine code.
    - He was furious when his students "wasted precious machine time" on assembly, considering it clerical work for humans.
  - Fortran [1957] [04:30]
    - John Backus and his team developed Fortran, a high-level language for mathematical notation and loops. [Image 18]
    - Assembly programmers were also uninterested, not understanding its benefits.
  - The Lesson: Resistance to Unlearning [04:52]
    - It's easy to assume technology always improves, but people often resist unlearning old ways and adopting new thought processes.
    - This resistance is a crucial factor when considering the future of programming.

Four Big Ideas for the Future of Programming [05:25] [Image 22]

1. From Coding to Direct Manipulation of Data [05:32]
- Today, programming involves writing lists of instructions (code).
- Ivan Sutherland's Sketchpad [1962] [06:47]
  - Sutherland's PhD thesis introduced Sketchpad, allowing direct drawing and manipulation of graphics on a video display. [Image 26]
  - Users could apply constraints (e.g., perpendicular lines) to sloppily drawn figures, and the system would automatically adjust them.
  - This created a "program" that draws a rectangle not through code, but by directly manipulating data and applying constraints.
  - Later, Sketchpad was used for complex tasks like bridge simulations, where constraints automatically handled physics. [Image 28]
  - Victor envisions a future where document creation, especially for visual or physical concepts, will be done through direct manipulation, not markup languages.
2. From Procedures to Goals and Constraints [09:43] [Image 31]
- Current programming uses procedures, instructing the computer "how" to do something.
- The future involves telling the computer "what" you want (goals), and it figures out "how."
- Carl Hewitt's Planner [1969] [10:19]
  - Planner, an early AI system, could reason both forward (procedurally) and backward (from goals). [Image 34]
  - Example: If Planner knows "apples are red," giving it an apple leads to "it's red." Asking for something red leads to "let's try an apple."
- Pattern Matching [11:45]
  - Pattern matching, with examples like SNOBOL [1962] and Regular Expressions [1967], is an important underlying concept that enables the goal-driven programming of Planner. [Image 37]
- Prolog [1972] [10:55]
  - Built upon Planner's backward reasoning, Prolog allowed programs to be expressed solely as goals.
  - This genre, called logic programming, emphasizes defining desired outcomes over step-by-step instructions.
- Communicating with Aliens: The Intergalactic Computer Network [1963] [12:53]
  - J. C. R. Licklider championed the idea of a global network of computers (the "Intergalactic Computer Network," later ARPANET/Internet). [Image 42]
  - A key challenge ("communicating with aliens") is how programs, written independently, can understand and interact with each other. [Image 47]
  - Directing programs with goals and constraints is the only scalable way for them to dynamically figure out how to communicate and fulfill shared objectives on a vast network, avoiding brittle API dependencies.
3. From Text Dumps to Spatial Representations [16:26] [Image 50]
- Today's programs are large text files, a sensible format for linear media like punch cards or paper tape.
- However, video displays enable new ways to represent dynamic information spatially.
- Doug Engelbart's NLS (Online System) [1968] [17:21]
  - Engelbart's NLS demonstrated displaying information on a video screen, with users pointing (using a mouse, a novel device) to get more details. [Image 54]
  - NLS introduced different "views" of information, presenting data as lists or two-dimensional diagrams.
- RAND Corporation's GRAIL [1968] [18:20]
  - GRAIL allowed programming using flowcharts on a video display with a stylus. [Image 58]
  - Users could draw shapes, and the system would recognize them as flowchart elements, assigning semantic meaning and allowing dynamic manipulation. [Image 60]
- Larry Tesler's Smalltalk Browser [1972] [19:35]
  - In Smalltalk, code is organized spatially in a browser, rather than long text files, for easier navigation and understanding. [Image 62]
  - Victor asserts that within 40 years, coding in text files will be obsolete, replaced by spatial representations.
- The Importance of Immediate Response [20:41]
  - Early interactive computing systems like NLS, GRAIL, Smalltalk, and PLATO (University of Illinois) prioritized immediate response in user interfaces. [Image 66]
  - This is crucial for simulating physical objects and creating intuitive, lag-free interactions.
4. From Sequential to Concurrent Programming [21:57]
- Current programming is predominantly sequential, reflecting the Von Neumann architecture (CPU fetching one word from memory at a time), which presents a "bottleneck" between CPU and memory. [21:57] [Image 73]
- This architecture is inefficient for modern needs, as most memory remains idle.
- The Microprocessor [1971] [23:33]
  - Intel's 4004 microprocessor, an entire processor on a single piece of silicon, revolutionizes hardware.
  - Future hardware will consist of "massively parallel processor arrays," vast networks of tiny, independent computers. [23:33] [Image 76]
- Limitations of Threads and Locks [25:38]
  - The current parallel programming model, using threads and locks, is inherently flawed and does not scale for these new architectures. [25:38] [Image 79]
  - It's difficult to reason about hundreds of threads competing for shared memory.
- Carl Hewitt's Actor Model [1973] [26:19]
  - Inspired by physics, the Actor Model views computation as independent "actors" with their own state, asynchronously sending messages to each other. [Image 85]
  - This model is a promising fit for massively parallel hardware, emphasizing independent action and communication over shared memory access.
  - This will require new programming models that fit the hardware, moving beyond sequential thinking and the limitations of threads and locks.

Conclusion: The Danger of Knowing What You're Doing [31:23]

Victor reiterates that the true message of his talk is not just about the specific ideas.
The most dangerous thought for a creative person is to believe they "know what they're doing," as this closes them off to new possibilities and prevents innovation. [Image 97]
He urges programmers to say, "We don't know what programming is. We don't know what computing is. We don't even know what a computer is."
Only by embracing this uncertainty can one truly be free to think anything and drive future advancements.