Our new Live Interview Series, Venturing In VC, premieres next week on Jan. 4, 2022. Venturing in VC is a live interview series where our host, Landon Campbell, interviews investors about how they broke into VC, their daily routines, views on the future, and lessons learned along the way. Who’s our first guest? None other than Turner Novak, Founder of Banana Capital — a founder-focused fund that invests anywhere from Pre-Seed to Pre-IPO with an average initial check size of $150,000-$300,000. Save your free seat to learn how Turner views the world and shares lessons he’s learned along his journey. Inside.com/vivc Venturing in VC is sponsored by SeedInvest — Fundraising isn’t easy, but it just got easier. Apply to start your raise on SeedInvest, or browse highly vetted startups to invest in. |
| |
 2022 prediction: Major tech firms and labs will continue to pour money into large language-generating models, making them more sophisticated, transparent, and efficient. It's possible that OpenAI will release GPT-4, its updated language model capable of generating human-like text and turning language into computer code. More: - According to Nvidia, there continues to be "hyper scaling of AI models leading to better performance, with seemingly no end in sight."
- DeepMind, Alphabet's AI research firm, just released a language-generating model that can match the performance of others up to 25x its size.
- The trend will likely continue. Future releases could result in new capabilities and a new class of applications based on pre-built large language models.
- More efficient large language models will transform gaming, search, chatbots, virtual assistance, and more.
GPT-4: - Based on previous model releases, GPT-4 is now on track for a 2022 or 2023 release.
- Some industry experts predict GPT-4 could have 100 trillion parameters, 500x times larger than GPT-3. More parameters generally equal a more sophisticated model.
- However, OpenAI CEO Sam Altman has said GPT-4 won't be much larger than GPT-3 but will use a lot more computing. It will be a text model as opposed to multi-modal.
- GPT-4 will also focus much more on coding through OpenAI's Codex, a machine learning tool that translates text into code, he noted.
- Meanwhile, a future GPT-5 might pass the Turing test but may not be worth the effort, Altman said.
Risks and more: - While transformative, building larger language models will require huge computational resources with major environmental impacts.
- It's likely that researchers will continue to focus more on building more energy-efficient model architectures and training paradigms.
- Future models will be trained on more than text data and may include data from audio, video, and images, similar to OpenAI's DALL-E, a version of GPT-3 that generates images from text.
- DeepMind has explored the ethical and social risks of large language models, concluding that more work is required to determine ways to mitigate these risks, including the spread of misinformation, discrimination, and harmful stereotypes.
| |
Prediction: More AI regulations are coming in the very near term. In a new report, Deloitte predicts that regulators will engage in more talks to rein in AI in 2022, though laws resulting from this may not be fully enacted until 2023 and beyond. More: - In 2022, the EU will continue discussing its draft AI Law, which has yet to be formally adopted, and China could propose some limited laws during its clampdown on Big Tech.
- The U.S. could continue proposing smaller bills that could rein in some aspects of AI, such as making algorithms more transparent, but other laws like facial recognition bans may take place only at the state and local levels, according to forecasts.
- The U.S. still trails the U.K., Canada, Germany, and China when it comes to AI regulation.
What could happen: - Deloitte forecasts different scenarios that will play out as countries adopt more AI rules over the next two years.
- For example, regulations could force some AI stakeholders in different countries to shut down their AI features or get fined.
- It's possible that the U.S., EU, and China will pass conflicting AI rules that make it difficult, if not impossible, for companies to comply.
- One country's regulations could emerge as the "gold standard," similar to the EU's GDPR, according to Deloitte.
AI Bill of Rights: - Two White House science advisers expressed the need for a "bill of rights" to govern AI systems and protect people against faulty, harmful, and biased systems.
- These include automated hiring tools, virtual assistants that fail to understand accents, facial recognition that leads to wrongful arrests, and biased health care algorithms that discriminate against people of color.
- A tech-focused "bill of rights" would clarify the freedoms and rights that AI and data-driven technologies should respect, the advisers argued.
- The Biden administration will explore this issue further in 2022.
| |
 A message from COIN CLOUD Welcome to Easy Buckets...where bitcoin meets basketball for a slam dunk opportunity. You could win your share of $100,000 in Bitcoin while watching some of today’s top basketball talent explore digital currency. Here’s why people are excited about the opportunity to Watch & Win with Easy Buckets: - This informational video series features Atlanta Hawks star Trae Young, New Orleans Pelicans guard Josh Hart, Las Vegas Aces forward A’ja Wilson, and Stanford Cardinal guard Haley Jones as they navigate the new and often confusing world of digital currency.
- Guiding them on their journey as series host is podcaster, publisher and recognized crypto expert Joe Pompliano.
- New episodes launch every Sunday starting December 26, 2021, with additional prizes and entry opportunities available each week (no purchase necessary to win).
- You could take home between $500 and $5,000 in Bitcoin as one of our weekly giveaway winners!
Ready to score? Enter now for a chance to win your share of $100,000 in Bitcoin. I want to win!
| |
Prediction: AI will continue moving from servers to the edge, where it will store local data on IoT and other devices to "learn" a customer's habits over time. Edge AI, which combines AI with edge computing, will address concerns about privacy from storing data in the cloud. It will make devices like appliances and smartphones more personalized, intelligent, and faster through lower latency. More: - So far, AI use cases on the edge have been largely in the computer vision field. Next year could see this extend out further to multimodal solutions and more use cases in healthcare, robotics, advertising, and elsewhere.
- MLOps, in particular, will help drive the flow of data to and from the edge for organizations rolling out AI.
- It's likely that IT departments will focus more on edge management and production in 2022.
- Computer scientists are also redesigning algorithms for use on local devices like phones and laptops.
- Thousands if not millions more servers are expected to be deployed at the edge next year, fueled by IoT, 5G, and video streaming.
- Gartner predicts that deep learning will be included in more than 65% of all edge use cases by 2027.
| |
Prediction: The chasm will grow wider between the U.S. and China when it comes to AI cooperation and innovation as the two continue to compete fiercely in the area. U.S. sanctions against Chinese AI and tech firms may continue, stifling connections between the private sector and academia in both countries. China will continue to turn more domestically to AI chips and AI technology. More: - China is now a "full-spectrum peer competitor" of the U.S. in artificial intelligence, according to a recent report from Harvard University's public policy institute.
- In 2022, it's possible that China could make strides and come closer to beating the U.S. in some areas of AI.
- For example, China graduates 4x as many STEM bachelor's students and will graduate 2x more STEM PhDs than the U.S. by 2025.
- Chinese speech technology firms are already beating U.S. firms in all languages, including English, the report found.
- The U.S. is still noticeably ahead in the number of "best paper" awards given out at major AI conferences, a trend that could continue next year.
- Meanwhile, China has already surpassed the U.S. in AI chip patents. As of Aug. 2020, Chinese AI chip designer Cambricon Technologies had 4x as many patents in its portfolio compared to Nvidia's.
- In 2022, Chinese domestic AI chipmaking will begin to catch up to the overseas market, where Nvidia and others still dominate, according to the Global Times.
| |
Other AI predictions for 2022: - SeedInvest is where high-growth startups go to fundraise online. Apply now to get started.*
- The idea of AI and human augmentation — when the two cooperate and work together on projects — will become more popular in 2022.
- Cybersecurity firms will pump up their investments in AI as online threats soar. AI/ML can handle more data and has faster response times.
- Companies will use more synthetic data to train their AI systems, gearing them for very specific use cases when there's a lack of available real-world data and datasets.
- Efforts to incorporate responsible and ethical AI standards will ramp up, but won't be enough to noticeably address AI's inherent biases.
- AI advancements in low-code and no-code will usher in more "citizen developers" that can use AI to generate code and create their own applications and programs, furthering AI democratization.
- Better AI, robotics, and automation will see the technologies take over more jobs in 2022 as the pandemic continues and supply chain demands persist.
- Homomorphic encryption and federated learning, which are methods of safeguarding data used in AI model creation, will grow at double-digit rates next year, Deloitte predicts.
- Venture capitalists will continue to invest a record amount of funding into AI startups, particularly those working in natural-language processing, like Hugging Face and Cohere.
- The boom in AI and tech-related M&As will continue into 2022, including the closure of Microsoft's purchase of Nuance Communications and Oracle's acquisition of Cerner.
- Trust is the community-powered corporate card. Learn from each other's spending patterns, get trusted advice, and increase your buyer. 100% free.
*This is sponsored content.
| |
Upcoming Events: - January 04 - Venturing in VC Episode #1 - featuring Turner Novak (Register Here)
- January 05 - DAOs and the intersection of tech & media with Kinjal Shah (Blockchain Capital) (Register Here)
- January 06 - Augmented and Virtual Reality with Timoni West (VP of AR/VR at Unity) (Register Here)
- January 11 - The Business of Interactive Gambling with Jon Kaplowitz (Penn) (Register Here)
- January 11 - Venturing in VC Episode #2 (Register Here)
- January 13 - Education within the Metaverse with Steve Grubbs (VictoryXR) (Register Here)
- January 25 - Need To Know NoCoder - Leon Young (Cogniss) (Register Here)
- January 27 - Understanding the Metaverse (Register Here)
- February 03 - Blockchain Gaming & Play-to-Earn Economies (Register Here)
- February 10 - The Business of Wellness (Register Here)
- February 17 - Customer Acquisition in a post iOS 14 world (Register Here)
- February 23 - Need to Know NoCoder: Derya Sousa (Kianda Technologies) (Register Here)
- February 24 - Alternative Investments (Register Here)
- March 03 - The $12B Appetite for Amazon Aggregators (Register Here)
| |
 |  | Beth is an Inside Analyst and a former public policy/investigative reporter. She has covered AI, VR, technology, and e-commerce for Inside. When she's not busy writing, she's hiking in Arizona's desert, playing pickleball, or trying out new Mexican food recipes. |   | | Editor | Aaron Crutchfield is based in the high desert of California. Over the last two decades, he has spent time writing and editing at various local newspapers and defense contractors in California. When he's not working, he can often be found looking at the latest memes with his kids or working on his 1962 Ford. | |
Enter now for a chance to win your share of $100,000 in Bitcoin with “Easy Buckets” from Coin Cloud. | |
Comments
Post a Comment