AI is evolving quickly, and new types of models keep showing up, which makes it hard to predict what’s next. However, David Cox, who leads AI model research at tech giant IBM (IBM), sees a promising trend where smaller language models (SLMs) are outperforming larger ones. More specifically, huge, powerful models are being shrunk by nearly 10 times every 6 to 9 months without losing their abilities. This means that these smaller models can run faster, use less energy, and fit into smaller devices. As Cox explains, “we can pack more into smaller packages,” which could make AI more widely available.
Elevate Your Investing Strategy:
- Take advantage of TipRanks Premium at 50% off! Unlock powerful investing tools, advanced data, and expert analyst insights to help you invest with confidence.
This shift to smaller models is also helpful for businesses. According to Abraham Daniels, a Senior Technical Product Manager for IBM’s Granite suite of foundation models, companies want AI that’s tailored to their own needs and data, and SLMs make that easier and more affordable. On top of that, new technology called activated low-rank adapters (LoRA) lets one AI model perform multiple tasks without switching to a different model.
Because of these changes, Daniels believes that the industry may be hitting a “scaling wall,” where simply making models bigger doesn’t help anymore. Instead, the next big step could be something called generative computing. This approach treats models more like software programs by using structured inputs, safety checks, and clear rules to guide how the AI works. Cox describes it as thinking of a model’s memory like a new kind of data, and the model itself as the processor. This could make AI more reliable and easier to control.
Is IBM a Buy, Sell, or Hold?
Turning to Wall Street, analysts have a Moderate Buy consensus rating on IBM stock based on six Buys, six Holds, and one Sell assigned in the past three months, as indicated by the graphic below. Furthermore, the average IBM price target of $287.25 per share implies that shares are trading near fair value.
